Beruflich Dokumente
Kultur Dokumente
Yi-Ruei Jhan
3D TCAD
Simulation
for CMOS
Nanoeletronic
Devices
3D TCAD Simulation for CMOS Nanoeletronic
Devices
Yung-Chun Wu Yi-Ruei Jhan
•
3D TCAD Simulation
for CMOS Nanoeletronic
Devices
Lg=10nm
FinFET
D
Fin G
STI
B
D
GAA NWFET
L g=10nm GAA
NW
S
123
Yung-Chun Wu Yi-Ruei Jhan
Department of Engineering and System Department of Engineering and System
Science Science
National Tsing Hua University National Tsing Hua University
Hsinchu Hsinchu
Taiwan Taiwan
v
vi Preface
This book demonstrates how to use the Synopsys Sentaurus TCAD 2014 version
for the design and simulation of 3D CMOS (complementary metal–oxide–semi-
conductor) semiconductor nanoelectronic devices, while also providing selected
source codes (Technology Computer-Aided Design, TCAD). Instead of the built-in
examples of Sentaurus TCAD 2014, the practical cases presented here, based on
years of teaching and research experience, are used to interpret and analyze sim-
ulation results of the physical and electrical properties of designed 3D CMOSFET
(metal–oxide–semiconductor field-effect transistor) nanoelectronic devices,
including Si, Ge, InGaAs FinFET, GAA NWFET, junctionless FinFET, tunnel
FinFET. In final chapter, also predicts the feasible options for silicon and
germanium FET of ultimate minimum dimensions.
The book also addresses in detail the fundamental theory of advanced semi-
conductor device design for the further simulation and analysis of electric and
physical properties of semiconductor devices. The design and simulation tech-
nologies for nano-semiconductor devices explored here are more practical in nature
and representative of the semiconductor industry, and as such can promote the
development of pioneering semiconductor devices, semiconductor device physics,
and more practically-oriented approaches to teaching and learning semiconductor
engineering.
The book can be used for graduate and senior undergraduate students alike,
while also offering a reference guide for engineers and experts in the semiconductor
industry. Readers are expected to have some preliminary knowledge of the field.
vii
Contents
ix
x Contents
7.5 Example 7.3 (3D n-Type TFET with Asymmetrical Gate) . . . . . . . 272
7.5.1 Descriptions of Motivation and Principle . . . . . . . . . . . . . . 272
7.6 Summary of This Chapter. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 278
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 278
8 Extremely Scaled Si and Ge to Lg = 3-nm FinFETs
and Lg = 1-nm Ultra-Thin Body Junctionless FET Simulation . . . . . 279
8.1 Foreword . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 279
8.1.1 Challenges of Sub-10-nm Technology Node . . . . . . . . . . . . 280
8.1.2 Material Selection for Sub-10-nm Technology Node . . . . . 280
8.2 Design Guideline of Sub-20-nm to 9-nm Gate Length Si FinFET
of Wine-Bottle Channel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281
8.2.1 Device Structure and Sub-20-nm FinFET Experimental
Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281
8.2.2 Simulation Results and Discussion . . . . . . . . . . . . . . . . . . . 281
8.3 Study of Silicon Lg = 3-nm Bulk IM, AC, and JL FinFET . . . . . . 283
8.4 Study of Germanium Lg = 3-nm Bulk FinFET . . . . . . . . . . . . . . . . 291
8.5 Study of Silicon and Germanium UTB-JL—FET
with Ultra-Short Gate Length = 1 and 3 nm . . . . . . . . . . . . . . . . . . 297
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 302
Appendix: Synopsys Sentaurus TCAD 2014 Version Software
Installation and Environmental Settings . . . . . . . . . . . . . . . . . . . . . . . . . . 305
About the Authors
Dr. Yung-Chun Wu received his B.S. degree in Physics from National Central
University in 1996, his M.S. degree in Physics from National Taiwan University in
1998, and his Ph.D. from National Chiao Tung University, Taiwan, in 2005. From
1998 to 2002, he was an assistant researcher at National Nano Device Laboratories,
Hsinchu, Taiwan, where he was primarily engaged in research on single electron
transistor and electron beam lithography technology. In 2006, he joined the
Department of Engineering and System Science, National Tsing-Hua University,
Hsinchu, Taiwan, where he is currently working as an associate professor. He
teaches 3D CMOS semiconductor nanoelectronic devices by TCAD simulation
course for ten years. His research interests include nanoelectronic devices and 3D
TCAD simulation, flash memory devices, and solar cells. He has published 56
international SCI papers on nanoelectronic devices.
Yi-Ruei Jhan received the B.S. degree in Physics from National Dong Hwa
University in 2010, M.S. degree in Engineering and System Science from National
Tsing Hua University in 2012, and Ph.D. degree in Engineering and System Science
from National Tsing Hua University in 2015. In 2016, he joined the Research and
Development department of Taiwan Semiconductor Manufacturing Company
(TSMC) after his graduation. His research interests include Nanoelectronic MOSFET
devices, TCAD simulation and Nonvolatile memory devices. He is author of book: 3D
TCAD Simulation for CMOS Nanoeletronic Devices.
xiii
Chapter 1
Introduction of Synopsys
Sentaurus TCAD Simulation
1.1 Introduction
current (Ion), OFF current (Ioff), operation voltage, threshold voltage (Vth), operation
frequency (f), inverter, and SRAM circuits. All semiconductor processes, device
parameters, and impacts on device properties can be analyzed by the design of 3D
semiconductor devices with different structures and materials. We can analyze the
strengths and weaknesses of 3D device’s key performance indicator (KPI).
This TCAD tool can greatly reduce the time and cost for R&D of top semicon-
ductor companies (such as TSMC, Intel, Samsung, IBM, UMC, Global foundry).
The purpose of this book is to not only allow readers to understand and use the
Synopsys Sentaurus TCAD 2014 version for design and simulation of 3D
semiconductor device based on integrated fundamental theory of semiconductor
device, but also simulate the electrical and physical properties of advanced 3D
semiconductor device in conjunction with the capability of software-aided design of
3D semiconductor devices.
This book emphasizes three major subjects: Part I (Chaps. 1–4) are about the
simulation of electrical properties of silicon CMOSFET, starting with the designs of
2D MOSFET and 3D silicon FinFET CMOS devices and circuits; Part II (Chaps. 5–7)
are about advanced nanoscale semiconductor devices such as Chap. 5: GAA NWFET,
Chap. 6: junctionless FET and Chap. 7: tunneling FET; Part III (Chap. 8) is about
predicting the feasible options for silicon and germanium FET of ultimate minimum
dimensions.
Instead of the built-in examples of Sentaurus TCAD 2014, the examples in this
book are the practical cases in years of our group research results and teaching
course in National Tsing Hua University, Hsinchu, Taiwan. The design and sim-
ulation technologies of nano-semiconductor device discussed in this book are rather
practical and representative in semiconductor industry and among academic
semiconductor researches. This book can help the development of advanced
nanoscale semiconductor device; understand semiconductor device physics, and the
practically learn the semiconductor engineering by TCAD simulation. This book is
suitable for the learning by graduated students who engaged COMS nanoeletronic
devices. It also can serve as the reference for engineers and experts in semicon-
ductor industry.
Moore’s law was proposed by Gordon Moore, one of the founders of Intel. It is the
observation that the number of transistors in a dense integrated circuit doubles
approximately every 18–24 months. This trend has continued for over half a cen-
tury. Moore’s law is actually a prediction of development of semiconductor
industry rather than a real law of physics. It is expected that Moore’s law is
expected to hold until 2030. According to 2015 International Technology Roadmap
for Semiconductors (ITRS) version 2.0 [2], the device miniaturization development
over the past few years is as shown in Fig. 1.1, and the R&D process of logic
device of Intel [3] is as shown in Fig. 1.2.
1.2 Introduction of Moore’s Law and FinFET 3
Fig. 1.1 Selected logic core device technology road map as predicted by 2015 ITRS version 2.0
[2]
Gate
Gate
Gate
Source Si Drain
Gate
Source Si Drain
Gate
Source
Drain
Fig. 1.2 Schematic plots of a FinFET, b fully depleted silicon-on-insulator FET, c vertical
nanowire gate-all-around FET, and d monolithic 3D FET, after 2015 ITRS version 2.0
With the world’s attention, Intel Developer Forum (IDF) [3] was held in San
Francisco, USA, 2015. Intel indicated in this forum that Moore’s law would con-
tinue to lead the breadth and speed of “innovation and integration” based on the
company’s technical advantages of nanoscale processes. In 2015, Intel introduced
the new-generation 14-nm Fin-shaped field-effect transistor (FinFET) CPU
Broadwell platform by adopting the advanced fabrication process of 14-nm
FinFET CPU together with the Intel second-generation 3D FinFET technology.
Intel is the first semiconductor company to enter the 14 nm era, and Broadwell CPU
will be the first to adopt this advanced process. The ultra-low voltage Core M series
customized for “Y” series CPU for ultra-slim tablet PC has been launched to the
market at the end of 2015. A part of details of 14-nm technology was publicly
disclosed by Intel in 2014 IDF: The thermal design power (TDP) of the new
product is only less than half of the previous generation, while it can provide similar
performance with better lifetime. Intel Broadwell structure has been optimized with
respect to the advantage of new feature of 14-nm process by adopting the
second-generation FinFET. It will be applied to various high-performance
1.2 Introduction of Moore’s Law and FinFET 5
Vdd (V)
Lg (nm)
18 LP 0.6
16
0.5
14
HP
12 0.4
10
8 0.3
2015 2018 2021 2024 2027 2030 2015 2018 2021 2024 2027 2030
Year Year
Fig. 1.3 Prediction plots of 2015 ITRS for a physical gate length (Lg) for HP and LP, b Fh and
Fw, c Vdd, and d Vtsat for HP at Ioff = 100 nA/lm and Vtsat for LP at Ioff = 100 pA/lm. HP
high-performance technology and LP low-power technology. Fh Fin height of FinFET, Fw Fin
width of FinFET
(a) (c)
1600 300
280
Ion, after Rext (uA/um)
1400 260
Rext (Ohms.um)
HP 240
1200 220
200
1000
180
800 160
LP 140
600 120
100
400 80
2015 2018 2021 2024 2027 2030 2015 2018 2021 2024 2027 2030
Year Year
200 3.5
180 3.0
140 2.0
120 1.5
100 1.0
80 0.5
2015 2018 2021 2024 2027 2030 2015 2018 2021 2024 2027 2030
Year Year
Fig. 1.4 Prediction plots of 2015 ITRS for a Idsat for HP at Ioff = 100 nA/lm and Idsat at
Ioff = 100 pA/lm, b effective mobility, c source/drain resistance, and d intrinsic delay (CV/I)
HK
Si Substrate Si Substrate
Fig. 1.5 Differences of shapes between 2015 Intel first-generation high-k metal gate (HKMG)
FinFET (or called tri-gate FET) and second-generation HKMG FinFET
1.3 Sentaurus Window Environment and Workbench for TCAD Task Management 7
60nm 42nm
pitch pitch
34nm 42nm
height height
Si Substrate Si Substrate
Fig. 1.6 Differences of Fin pitches and heights of 2015 Intel first-generation 22-nm FinFET and
second-generation 14-nm FinFET
Fig. 1.7 Synopsys Sentaurus TCAD is a complete graphical operating environment which
includes numerous simulation tools (Copyright © Synopsys, Inc. All rights reserved.)
tools, setting of important variables, and planning process flow for a project. The
simulation results can be presented in the form of visual display. The simulation
raw data can also be exported via proper graphical analysis software for analyzing
electrical and physical properties (Fig. 1.7).
8 1 Introduction of Synopsys Sentaurus TCAD Simulation
Fig. 1.9 Typical tool flow with device simulation using Sentaurus Device
Sentaurus Structure
SNMESH SDEVICE INSPECT
Workbench Editor
Fig. 1.10 Basic process flow charts of simulation tools by Synopsys Sentaurus TCAD 2014
version and the required simulation tool software
4. Descriptive command bar which can be accessed and recorded from graphical
interface.
In this book, we start from SDE tool to establish 3D nanoelectronic device
structure by defining several “blocks.” With given device’s dimension, materials,
and different dopant of each block, complicated structures such as FinFET or
GAA FET can be easily created by arrangement and combination of these blocks.
SDE tool also allows definition of variable parameters for subsequent adjustment on
SWB, such as thickness of gate oxide, length of gate, metal work function, and
operating voltages. Other important semiconductor technologies, such as silicide,
high dielectric materials, metal gates, lightly doped drain (LDD), and body bias, can
also be easily designed and simulated in Synopsys Sentaurus TCAD.
For example, the FinFET structure based on silicon bulk is shown in Fig. 1.12.
(3) SNMESH
SNMESH tool refers to the points of mathematic model to be solved, where the
density of mesh can be self-defined. The location with denser mesh can better
reflect the variation of physical properties of this area, such as potential gradient,
electric field gradient, and carrier concentration gradient. Excessive mesh will result
in prolonged simulation time.
For example, the mesh on FIN structure of Bulk FinFET is shown in
Figs. 1.13 and 1.14.
Fig. 1.12 FinFET structure on bulk is established by the permutation and combination of 3D
blocks
12 1 Introduction of Synopsys Sentaurus TCAD Simulation
(4) SDEVICE
SDEVICE tool is a general-purpose device simulation tool which offers simulation
capability in the following broad categories:
(a) Advanced Logic Technologies: Sentaurus Device simulates advanced logic
technologies such as Si FinFET and FDSOI, including stress engineering,
channel quantization effects, hot carrier effects and ballistic transport, and many
other advanced transport phenomena. Sentaurus Device also supports the
modeling of SiGe, SiSn, InGaAs, InSb, and other high-mobility channel
materials and implements highly efficient methods for modeling atomistic and
process variability effects.
(b) Compound Semiconductor Technologies: Sentaurus Device can simulate
advanced quantization models including rigorous Schrödinger solution and
complex tunneling mechanisms for transport of carriers in heterostructure
devices such as HEMTs and HBTs made from, but not limited to, GaAs, InP,
GaN, SiGe, SiC, AlGaAs, InGaAs, AlGaN, and InGaN.
(c) Optoelectronic Devices: Sentaurus Device has the capability to simulate the
optoelectronic characteristics of semiconductor devices such as CMOS image
sensors and solar cells. Options within Sentaurus Device also allow for rigorous
solution of the Maxwell’s wave equation using FDTD methods.
(d) Power Electronic Devices: Sentaurus Device is the most flexible and advanced
platform for simulating electrical and thermal effects in a wide range of power
devices such as IGBT, power MOS, LDMOS, thyristors, and high-frequency
high-power devices made from wide bandgap material such as GaN and SiC.
(e) Memory Devices: With advanced carrier tunneling models for gate leakage and
trapping de-trapping models, Sentaurus Device can simulate any floating gate
device like SONOS and flash memory devices including devices using high-k
dielectric.
(f) Novel Semiconductor Technologies: Advanced physics and the ability to add
user-defined models in Sentaurus Device allow for investigation of novel
structures made from new material.
(5) INSPECT
INSPECT tool is used for extracting current and voltage properties of semicon-
ductor device, such as:
1. Subthreshold swing (SS).
2. Threshold voltage (Vth).
3. Drain-induced barrier lowering (DIBL).
4. Transconductance (Gm).
5. Saturation current (Isat).
6. Off-state leakage current (Ioff).
7. Resistance (Rout)
14 1 Introduction of Synopsys Sentaurus TCAD Simulation
8. Inverter performance
9. SRAM performance
10. Analog/RF performance.
The SWB family tree view of simulation project is as shown in Fig. 1.15 with
user-friendly window-based user interface.
Sentaurus TCAD Toolbar Buttons of user-friendly window-based user inter-
face are shown in Fig. 1.16.
Sentaurus Visual tool is the advanced visualization software for TCAD data
analysis. It is equipped with rich graphics capabilities for interactive composition of
X–Y curves and 2D/3D TCAD device structures and device electrical and physical
properties. The 2D and 3D user interfaces of Sentaurus Visual are shown in
Figs. 1.17 and 1.18.
In addition, semiconductor device technology integrated with virtual process is
user-friendly. Semiconductor device simulation technologies such as Front End of
Line (FEOL) and Back End of Line (BEOL) can all be processed by tools such as
Sentaurus Interconnect. The strong mathematical simulation algorithm is capable of
simulating technical steps of ion implantation, thermal diffusion, doping activation,
3D – nFinFET Mesh
L g =15nm
V d =1V
V g =1V
Calibration and consulting services are also available by Synopsys company sup-
port. Currently, all forefront technologies and device engineers of semiconductor
companies are using Synopsys TCAD tool to develop and optimize forefront
semiconductor device technologies. Sometimes, these tools need to be calibrated
with respect to specific technology in order to improve the predictability of future
nodes or special process. Synopsys TCAD provides customers with calibration,
1.7 Calibration and Services 17
simulation, model development, and integration services. The calibration and ser-
vices provided by Synopsys based on complete and verified TCAD technology and
TCAD device module solutions can accelerate technological development and
solve problems of fabrication process of forefront semiconductor companies, thus
enhancing the international competitiveness of semiconductor company.
References
1. Synopsys Sentaurus TCAD Ver. J-2014.09, Synopsys, Inc., Mountain View, CA, USA.
(2014). https://www.synopsys.com/home.aspx
2. ITRS version 2.0. (2015). http://www.semiconductors.org/main/2015_international_
technology_roadmap_for_semiconductors_itrs/
3. Intel Developer Forum on line material, San Francisco, CA, USA. (2014)
Chapter 2
2D MOSFET Simulation
0V Vdd
Fig. 2.1 a N-channel FET is in off-state, b N-channel FET is in on-state, c P-channel FET is in
off-state, d P-channel FET is in the on-state [1]
pulled up to Vdd. Figure 2.2b shows how nFET and pFET are fabricated on the same
chip. An N-type well is formed in a portion of area in the P-type silicon substrate by
implantation or diffusion of N-type dopant. The contacts of P-type silicon substrate
and N-type well are both shown in the figure. The layout of basic CMOS inverter is
as shown in Fig. 2.2c. This is an image of top–down view of silicon water, which is
also the image composed of several overlapped masks for fabrication of inverter. Vin,
Vout, Vdd, and ground point voltage are all based on metal lines. Polysilicon gate or
metal gate is the vertical line connected to Vin.
In this chapter, the physical and electrical properties of MOSFET will be
investigated by 2D TCAD simulation. It is shown in Fig. 2.2 that the current is
flowing from the high potential terminal to the low potential terminal.
The current (Ids) in Fig. 2.3 can be derived from the charge concentration and
width of inversion layer as shown below:
ZL ZVds
Ids dx ¼ WCoxe lns ðVgs Vcs Vt ÞdVcs ð2:2Þ
0 0
2.1 Complementary MOS (CMOS) Technology 21
pFET
nFET
Gate Vout
0V
(b) Vin
Vin
0V Vout Vdd
P
+
N
+
N
+
P
+
P
+
N
+ nFET
N-well 0V
P-substrate
Fig. 2.2 Three figures of CMOS Inverter, a CMOS inverter circuit consists of a pFET pull-up
device and a nFET pull-down device, b nFET and pFET device structures are integrated in a same
chip, and c the layout of CMOS inverter
Gate
Tox
Ids
Vs N+ N+ Vds
Wdmax
Vb
0 L
22 2 2D MOSFET Simulation
1
Ids L ¼ WCoxe lns Vgs Vt Vds Vds ð2:3Þ
2
Ids ¼ WL Coxe lns Vgs Vt 12 Vds Vds ð2:4Þ
Equation (2.4) reveals that Ids is proportional to W, lns, Vds/L (average electric field
in the channel), and Cox ðVg Vt 12 Vds Þ. The Cox ðVg Vt 12 Vds Þ can reflect the
inversion electron density Qinv (C/cm2) in the channel. Coxe is effective gate
capacitance (F/cm2).
When Vds is very small, the item of 1/2 Vds of (2.4) can be neglected, so Ids is
proportional to Vds, which means the conduction behavior of transistor under such
voltage making it just like a resistor. The I–V When Vds is increased, Qinv, and
dIds =dVds are decreased. The differentiation of Eq. (2.4) with respect to Vds will
result in dIds =dVds , which will be equaled to 0 under a specific Vds. At this moment,
the Vds is called Vdsat as shown below:
dIds W
¼ 0 ¼ Coxe lns ðVgs Vt Vds Þ; when Vds ¼ Vdsat ð2:5Þ
dVds L
Vdsat is called drain saturation voltage, and different Vgs will lead to different Vdsat.
The region with Vds far less than Vdsat is called linear region, and the region with Vds
greater than Vdsat is called “saturation region.” The part of I–V curve of Fig. 2.4
with Vds Vdsat is the “linear region.”
Vgs=1.5V
Ids (A)
5e-6
Vgs=1.0V
Vgs=0.5V
0
0 0.5 1.0 1.5
Vds (V)
2.1 Complementary MOS (CMOS) Technology 23
The current in saturation region can be derived from Eq. (2.4) as:
Idsat ¼ 2L
W
Coxe lns ðVgs Vt Þ2 ð2:7Þ
W
gmsat ¼ Coxe lns ðVgs Vt Þ ð2:9Þ
L
The N-type 2D nMOSFET with different Lg = 200, 400, 600, 800, 1000 nm is used
as an example for the introduction of simulation technology. First, the Synopsys
Sentaurus TCAD 2014 version is used to establish the four tools of SDE,
SNMESH, SDEVICE, and INSPECT as shown in Fig. 2.5.
Fig. 2.5 Required simulation four tools are shown in the workbench of 2D MOSFET simulation
24 2 2D MOSFET Simulation
y 2D MOSFET (N channel)
Unit nm
25nm 25nm @Lg@ 25nm 25nm
y3
@tox@ SiO2
y2
@tac@ P 1E20 P 1E20 B 1E16 P 1E20 P 1E20
(000)
Silicon
Body B 1E15
y1
x
x1 x2 x3 x4 x5
Contact
Now, we start to establish the device structure. The commands icon on SDE tool
should be right-clicked to enter codes to establish the structure. At first, we need to
draw a cartoon plot as shown in Fig. 2.6 of 2D MOSFET structure for simulation.
The structure of 2D MOSFET to be established is as shown in Fig. 2.6.
1. SDE ! devise_dvs.cmd
Now, we use Fig. 2.6 to explain the SDE tool commands (code file is devise_dvs.
cmd)
In principle, the flow of structure establishment is as shown below:
(1) Set zero point and coordinates
(2) Composition of 2D Structure
(3) Composition of 3D Structure and 2D Y-cut diagram (in Chap. 3)
(4) Set 2D rectangles or 3D cuboids (in Chap. 3)
(5) Set Electrodes
(6) Set doping region
(7) Set Mesh
The codes of devise_dvs.cmd can be divided into six parts.
(1) Parameter
(2) Structure
(3) Contact
(4) Doping
(5) Mesh
(6) Save
2.2 [Example 2.1] 2D n-Type MOSFET Id–Vg Characteristics Simulation 25
(define x1 LSDC)
(define x2 (+ x1 LSD))
(define x3 (+ x2 Lg))
(define x4 (+ x3 LSD))
(define x5 (+ x4 LSDC))
(define y1 (- Body))
(define y2 tac)
(define y3 (+ tac tox))
;----- (2). Structure -----;
"ABA"
;--- source ---
(sdegeo:create-rectangle
(position 0 0 0)
(position x1 y2 0) "Silicon" "SourceC" )
(sdegeo:create-rectangle
(position x1 0 0)
(position x2 y2 0) "Silicon" "Source" )
;--- Channel ---
(sdegeo:create-rectangle
(position x2 0 0)
(position x3 y2 0) "Silicon" "Channel" )
26 2 2D MOSFET Simulation
;----- (6). Save (BND and CMD and rescale to nm) -----;
(sde:assign-material-and-region-names (get-body-list) )
(sdeio:save-tdr-bnd (get-body-list) "n@node@_nm.tdr")
(sdedr:write-scaled-cmd-file "n@node@_msh.cmd" nm)
28 2 2D MOSFET Simulation
(define sde:scale-tdr-bnd
(lambda (tdrin sf tdrout)
(sde:clear)
(sdegeo:set-default-boolean "XX")
(sdeio:read-tdr-bnd tdrin)
(entity:scale (get-body-list) sf)
(sdeio:save-tdr-bnd (get-body-list) tdrout)
)
)
(sde:scale-tdr-bnd "n@node@_nm.tdr" nm "n@node@_bnd.tdr")
;-------------------------------------------- END ---------------------------------------;
In this example, the thicknesses of gate length (Lg) and gate oxide (tox) layer are
set as variables @Lg@ and @tox@, each of which is composed by two rectangles.
The body is composed of one rectangle, so it is composed of a total of “seven 2D
rectangles.”
In the same line of code, materials, number or parameters with different definitions
must be separated by blanks. The universal code for parameter definition is (define A
B), in which A represents the name declared to the computer, and B represents the
value of A. The default unit of this simulation software is µm, but the unit of nm is
more suitable for current semiconductor device design. Therefore, the unit of nm
must be set in the parameter setting with the value of µm multiplied by 10−3, defined
as “(define nm 1e-3).” In addition, in this book, “tac” is defined as the thickness of
active layer, “tox” is defined as the thickness of gate oxide, “Body” is defined as the
thickness of body beneath active layer, “LSDC” is defined as the length of
source/drain contact, “LSD” is defined as the length of source/drain, “C_Doping” is
defined as the doping concentration of channel, “D_Doping” is defined as the doping
concentration of drain, “S_Doping” is defined as the doping concentration of source,
and “B_Doping” is defined as the doping concentration of body.
And then the algebra of x1, x2, x3… is used for representing all values to easily
facilitate the code adjustment during lots of parameter modification in the future and
the device structure scaling. It is revealed in Fig. 2.6 that the algebra of every
coordinate (x1, x2, …) is the sum of the previous algebra and the previously defined
parameter. For example, x2 is equaled to x1 plus LSDC.
It is noteworthy that, in this Sentaurus TCAD software, the form of universal
codes of addition/subtraction/multiplication/division is (operator A B). Its mathe-
matical meaning is the operator calculation of A with respect to B. For example, (+
10 5) means 10 + 5, (− 10 5) means 10 − 5, and (− 10 5 2) means 10 – 5 − 2.
Attentions must be paid to the sequence of multiplication and division operation.
For example, (/10 5 2) means (10/5)/2, and it is because the order of calculation in
this program is from left to right.
If there is another parenthesis behind 10, the calculation order can be changed.
For example, (/10 (/5 2)) means 10/(5/2). And then, we enter the section of device
structure establishment.
In this example, the device structure is established via rectangle stacking. It is
shown in Fig. 2.6 that a total of seven rectangles are used in this example. The
benefit of this method is that it can easily establish the desired structure. However,
there will be the problem of computer failed to identify the intersection or over-
lapped region of rectangles. “ABA” is exactly the code to deal with this issue. It
means that the new rectangle will replace the old rectangle at the region with
overlapping new and old rectangles, and the material property of this region will be
determined by the new rectangle as shown in Fig. 2.8.
As for the device structure establishment, the establishment of rectangles will be
declared by sdegeo. The rectangle refers to the 2D rectangular rectangle, and the
size of rectangle is determined by the diagonal.
As shown in Fig. 2.9, the size of rectangle can be determined by the assigned
coordinates of A and B. For example, (sdegeo:create-rectangle (position 0 0 0)
30 2 2D MOSFET Simulation
(b)
Origin Box (Gox)
Origin Origin
New New
box box
box
Box
(Si)
Contact code: define-contact-set “G” is for assigning the name of contact which
can be self-defined. (color:rgb 1.0 0.0 0.0) “##” is for defining the color and form
of contact, and here it can be set as default. This example is a 2D simulation, so the
2D contact needs to be established such as define-2d-contact. It is noteworthy that,
as described in previous section, only the electrode equipotential lines need to be
defined (equipotential surface in 3D structure), and there is unnecessary for
defining physical material and physical space of contact. By assigning a point to
the defined equipotential line, the program will automatically stretch to the left and
to the right from that point until reaching the boundaries of this 2D rectangle, and
this extended line is the closed electrode equipotential line. The code is
find-edge-id (position (+ x2 (/ Lg 2)) y3 0)) “G”. It is noteworthy that the name of
electrode must be given once here, and the definition G means “gate contact
electrode.” And then, the contact electrodes of source, drain, and substrate must be
established in proper order.
After all required contacts are established, the next thing is to define the doping
styles and doping concentrations of seven rectangles.
The rule of doping establishment is to determine the style and concentration of
doping before putting into the previously established rectangles to complete the
doping process. The code of “define-constant-profile” refers to a doping with
fixed style and concentration which does not take into consideration the concen-
tration gradient.
“dopedC” refers to the names of this doping style and concentration, which can
be assigned based on personal preference. “BoronActiveConcentration” refers to
a dopant which is already activated, and here it is boron. The C_Doping refers to
the concentration of this doping, and the value of C_Doping has been assigned
during parameter definition.
Code: define-constant-profile-region refers to putting the pre-defined doping
style and concentration into a fixed region without considering the situation of
dopant diffusion. “RegionC” refers to the name of this action which can be set in
accordance with personal reference. The content of this action is to combine all
codes showing up afterward, so if “RegionC” is followed by “dopedC” and
“Channel,” it refers to putting dopedC into the rectangle of Channel, and this
action is called RegionC. This is how the doping of seven rectangles is composed.
After the definition of doping is completed, the mesh of mathematical calcula-
tion should be determined. The simulation calculation of semiconductor device
must be analyzed by various physics formula, the most fundamental of which are
the Poisson equation to determine electrical potential and the continuity equation to
determine carrier concentration. The electrical properties of semiconductor device
must be based on the simultaneous solution, such as Poisson equation, continuity
equation, and transport equation. The location of such solution is the intersection of
mesh as shown in Fig. 2.10. However, there cannot be infinite number of solutions
for semiconductor device, so Newton interpolation method is used for approxi-
mation between points. However, the approximation by interpolation is not a proper
32 2 2D MOSFET Simulation
Fig. 2.10 Actual example of mesh. The active layer is designed with denser mesh, and the
substrate layer is designed with less dense mesh in order to obtain optimized simulation design
efficiency
option for regions with large concentration gradient or electric field variation. These
regions must be analyzed by more precise solutions, so the meshes in these regions
must be denser.
The logic behind the establishment of mesh is: First a comprehensive mesh is
assigned. Since the locations of solutions are the intersections of mesh, a dense
mesh will be assigned to the region with large concentration gradient, large electric
field variation, or significant impact on electrical properties (such as the active
region). Therefore, an AllMesh is assigned first. define-refinement-size refers to
the definition of distance between points. With varying geometric shapes of
semiconductor device, there will be different distributions of dopant concentrations
such that the program will automatically adjust the distance on the boundary as long
as the maximum and minimum values are declared to the computer. “Cha_Mesh”
refers to the name of aforementioned action, which can be determined based on
personal preference. Among the next six numbers, the first three of them refer to the
maximum values along the three directions of X-axis, Y-axis, and Z-axis, and the
next three of them refer to the minimum values along the three directions of X-axis,
Y-axis, and Z-axis. In this case, these numbers are (20 20 0 10 10 0), because the
maximum value of X is 20 nm and the minimum value of X is 10 nm, and the same
shall apply to Y, so the maximum and minimum values of Z are set to be zero,
because this example is 2D device.
2.2 [Example 2.1] 2D n-Type MOSFET Id–Vg Characteristics Simulation 33
The mesh established earlier can be placed in specific region or specific material,
and here it is placed in the specific material by the code: define-refinement-
material. “channel_RF” is the name of this action, which can be determined based
on personal preference. However, it is better to be in accordance with the syntax
suggested in this book to avoid unnecessary error. This action is to combine the
following code, so if “channel_RF” is followed by “Cha_Mesh” and “Silicon,” it
refers to the action of placing Cha_Mesh into the material of silicon. This action is
called channel_RF (i.e., channel refinement).
After the comprehensive mesh is established, a denser mesh will be estab-
lished in the active layer. The method for doing this is to establish a mesh
rectangle and then put it in the designated region. The code: define-refinement-
window is for establishing the mesh rectangle. The name of this action is
multiboxChannel, which can be determined based on personal preference, yet it is
better to be in accordance with the syntax suggested in this book to avoid unnec-
essary error. “Rectangle” refers to the establishment of a 2D mesh rectangle based
on the method of determining the size of rectangle by diagonal before assigning the
maximum and minimum values along the three directions of X-axis, Y-axis, and
Z-axis. And the code: multiboxPlacementChannel is the name of this action
which can be determined based on personal preference. This action is to integrate
the following code, so if “multiboxPlacementChannel” is followed by
“multiboxSizeChannel” and “multiboxChannel,” it refers to placing
multiboxSizeChannel into multiboxChannel, and this action is called
multiboxPlacementChannel.
The denser mesh will be stretched across the region with larger variation of con-
centration gradient, so here another line of code will be added for the denser mesh to be
stretched toward the depletion region. The code is define-refinement-function. This
action will take place in multiboxPlacementChannel with extension in accordance
with DopingConcentration. The MaxTransDiff refers to the degree of extension,
which is one. There has more selection of MaxTransDiff degree; readers can refer the
TCAD manual.
And next the file is saved with the code as shown in the last part of code:
The part can be used as default. Among them, assign-material-and-region-names
is for saving the previously established materials and names, and the line of
write-scaled-cmd-file is for saving the aforementioned scale. With the default unit
of program being µm, X-axis, Y-axis, and Z-axis must be multiplied by 10−3 and
the unit should be converted to nm. The following scale-tdr-bnd saves all files as
boundary format. It will be used by the following tool. All codes required by SDE
are hereby completed.
The second tool in the tool column is “SNMESH” tool in SWB, which is mainly
used for establishing the mesh required by device simulation, and the mesh code
has been written in the commands of SDE. So we only have to set SNMESH to
access the commands of SDE as shown in Fig. 2.11.
34 2 2D MOSFET Simulation
NaturalBoxMethod
}
Plot{
eDensity hDensity
eCurrent hCurrent
TotalCurrent/Vector eCurrent/Vector hCurrent/Vector
eMobility hMobility
eVelocity hVelocity
eEnormal hEnormal
ElectricField/Vector Potential SpaceCharge
eQuasiFermi hQuasiFermi
Potential Doping SpaceCharge
SRH Auger
AvalancheGeneration
DonorConcentration AcceptorConcentration
Doping
eGradQuasiFermi/Vector hGradQuasiFermi/Vector
eEparallel hEparalllel
BandGap
BandGapNarrowing
Affinity
ConductionBand ValenceBand
eQuantumPotential
}
Solve {
Coupled ( Iterations= 150){ Poisson eQuantumPotential }
Coupled { Poisson eQuantumPotential Electron Hole }
Quasistationary(
InitialStep= 1e-3 Increment= 1.2
MinStep= 1e-12 MaxStep= 0.95
Goal { Name= "D" Voltage=@Vd@ }
){ Coupled { Poisson eQuantumPotential Electron Hole } }
Quasistationary(
InitialStep= 1e-3 Increment= 1.2
MinStep= 1e-12 MaxStep= 0.02
2.2 [Example 2.1] 2D n-Type MOSFET Id–Vg Characteristics Simulation 37
At first, all the conditions of voltage (Vd, Vs, and Vg) will be defined together
with the gate. The metal gate work function (WorkFunction(V)) sets as a variable
@WK@.
The name of each electrode terminal must be identical to the name in SDE
tool, or the program cannot be executed due to interpretation failure. The work
function will be set as variables to be defined in workbench to facilitate the cali-
bration of threshold voltage (Vth).
The next step is to set up the files to be read from the previous tool during the
operation of SDEVICE. This part is based on default value such that it cannot be
modified without permission, because the file name must be in compliance with the
program regulation.
And then, we need to tell the computer what are the physics formula to be
substituted into this simulation calculation, such as the recombination model or
some quantum modification models.
Among them, the eQuantumPotential is the quantum modification item with
respect to the density of state of electron. When the device dimension is very small,
some quantum modification items must be added for the simulation results to be
closer to the real condition. The default setting of this program is based on the most
complete physics model, so please read the manual thoroughly before making any
modification.
After assigning physics model, the next step is the assignment of mathematical
model, which can be based on default setting. The default setting of this program is
based on the most complete mathematic model, so please read the manual thor-
oughly before making any modification.
It is noteworthy that Iterations = 20, it indicates the number of points to be
given during approximation by Newton interpolation method between two mesh
points, and 20 points are given here. Usually the less dense mesh will be set up first
before the mesh of important simulation step and crucial region (such as active
layer) can be optimized during simulation process in order to be in compliance with
the correct electrical and physical properties of the device. Therefore, this part of
38 2 2D MOSFET Simulation
#................................................................ #
puts $log_file "Threshold voltage VT1 = $VT1 Volts"
puts $log_file " "
#...................................................................... #
# 3) Initialization and display of curves on the main Inspect screen : #
# ..................................................................... #
cv_display idvgs
cv_lineStyle idvgs solid
cv_lineColor idvgs red
# ---------------------------------------------------------------------- #
# II) gm = maxslope((ID[VGS]) #
# ---------------------------------------------------------------------- #
set gm [ f_gm idvgs ]
puts $log_file " "
puts $log_file "Transconductance gm = $gm A/V"
puts $log_file " "
set ioff [ cv_compute "vecmin(<idvgs>)" A A A A ]
puts $log_file " "
puts $log_file "Current ioff = $ioff A"
puts $log_file " "
set isat [ cv_compute "vecmax(<idvgs>)" A A A A ]
puts $log_file " "
puts $log_file "Current isat = $isat A"
puts $log_file " "
set rout [ cv_compute "Rout(<idvgs>)" A A A A ]
puts $log_file " "
puts $log_file "Resistant rout = $rout A"
puts $log_file " "
cv_createWithFormula logcurve "log10(<idvgs>)" A A A A
cv_createWithFormula difflog "diff(<logcurve>)" A A A A
set sslop [ cv_compute "1/vecmax(<difflog>)" A A A A ]
puts $log_file " "
puts $log_file "sub solp = $sslop A/V"
puts $log_file " "
### Puting into Family Table #####
2.2 [Example 2.1] 2D n-Type MOSFET Id–Vg Characteristics Simulation 41
ft_scalar VT $VT1
ft_scalar gmax $gm
ft_scalar ioff $ioff
ft_scalar isat $isat
ft_scalar sslop $sslop
ft_scalar rout $rout
close $log_file
#--------------------------------- END --------------------------------------#
As for the extraction of Vth, the drain current is to be collected with this example
based on nFET, and it is set as D. If the simulation is for pFET, it should be
changed to S. As for the extraction of SS, if it is for pFET, diff(<logcurve>) must be
multiplied by (−1) before the current of pFET is in opposite direction to the current
of nFET.
By now all codes have been entered, and in the end, all parameters of workbench
should be set, and please remember to enter all variables to be included in the
workbench as shown in Fig. 2.12. The setting method is as shown in Fig. 2.13.
First, right-click to select Add in the lower column of Tool, and then enter the
name of variable in Parameter and the value of variable in Default Value. The
values can be deleted or changed by right-clicking Edit Values in the lower column
of Tool as shown in Fig. 2.14.
After the variables are completed, the project name can be selected, and run can
be clicked to start the simulation calculation as shown in Fig. 2.15.
It is shown in Fig. 2.16 that the name of project will turn yellow after the
completion of simulation, indicating a successful simulation calculation; if it turns
red, it means the simulation cannot be converged or has some syntax error. User
can check error message in log file. Also the mesh calculation may need to be
adjusted for convergence.
Fig. 2.12 Important variables being set up of each tool in the SWB
42 2 2D MOSFET Simulation
After the completion of simulation, the next step uses INSPECT tool to analyze
electrical properties as shown in Fig. 2.17, where the node of the electrical
property should be right-clicked and the eye icon on the visualize bar should be
clicked to select Inspect (All Files).
The selected interface is as shown in Fig. 2.18. The node to be inspected should be
selected on the datasheet of workbench, and the set electrode and the electrical
property to be inspected should be selected before being put on the preset axis. For
operation example: G ! OuterVoltage ! To X-Axis; D ! TotalCurrent ! To
Left Y-Axis. In the end, the Y-axis on the topmost tool bar should be changed to be
displayed in log scale to lead us to the frequently seen Id–Vg curve.
In addition, the data of Id–Vg curve can be exported in txt file format to be
analyzed by other professional engineering and scientific application graphics
software such as SigmaPlot and Origin. As indicated in Fig. 2.19, the data of
44 2 2D MOSFET Simulation
Fig. 2.18 Using Inspect tool plots electrical properties display and analysis. The plot shows a
typical Id–Vg transfer curve of nMOSFET
electrical properties can be exported by clicking the File ! Export ! csv or txt
format on the upper left corner. The data of electrical properties is exported in the
format of csv or txt file as shown on the right side of Fig. 2.19, where the data of x
is Vg, and the data of y is Id.
Except for the electrical property diagram, other important physical properties,
such as electric field, electrostatic potential, and charge concentration, need to be
examined during the analysis of semiconductor device. As shown in Fig. 2.20, the
node of electrical property to be examined should be right-clicked, and the eye icon
on the Visualize bar should be selected to access the drop-down menu of visual-
ization software. And then, we should select Sentaurus Visual (All Files).
2.2 [Example 2.1] 2D n-Type MOSFET Id–Vg Characteristics Simulation 45
Fig. 2.19 Exported data of electrical properties is in the txt file format (ex: IdVg.txt)
Figure 2.21 is the Sentaurus Visual interface. The Selection on the left is for
selecting the material to be displayed and its mesh. In addition, the physical
property to be inspected can be selected on the lower part of screen such as: energy
band diagram and carrier distribution of electrons and holes. If either X-axis or
Y-axis is to be fixed to observe the variation of physical property along with the
other axis, the icons in the red frame on the right can be selected. For example,
fixing X-axis at the position of 0.5 will allow us to observe the variation of physical
property along Y-axis with X = 0.5.
Special Note: FAQ and Troubleshooting
The most frequently seen problem is that the value does not converge during the
simulation, and the error message is as shown in Fig. 2.22.
Most of these problems are due to the difficulty in mesh calculation which has
resulted in singular points generation at the location with large concentration
variation gradient thus causing diversion. The better mesh code which is less
vulnerable to diversion as shown below:
2.2 [Example 2.1] 2D n-Type MOSFET Id–Vg Characteristics Simulation 47
Fig. 2.22 Error message of the value of simulation does not converge
The following fine cutting is mainly placed inside the channel because there is
large carrier variation gradient and large electric field variation. So the cutting
should be 1 nm ! 0.5 nm or 1 nm ! 2 nm. In this book, it is suggested that mesh
variation should be kept within 100%.
The max value and min value of mesh should be determined in coordination
with device dimension from high to low. The divergence will most likely to take
place in the region with large concentration gradient and electric field variation.
This is because the Newton interpolation method is used for approximation at the
intersection of mesh, and the value is confirmed by left limit and right limit
approach. If the values of left limit and right limit do not match, the simulation
result will diverge. Therefore, smaller mesh should be assigned to the location with
large concentration gradient or large electric field variation. The device for the first
run can be assigned with a larger mesh or fewer elements to see if the electric
properties are as expected and finer segmentation can be applied for observation of
electric field distribution or carrier distribution. It still cannot converge after mesh
adjustment, fine-tuning of workfunction can be considered. For example, if the
original value is 4.6 eV, we can try with 4.601 eV. An additional 0.001 eV will
not lead to too much impact on Vth, but it can help with convergence. Voltage
fine-tuning can also be applied such as changing VD = 1 V to VD = 1.01 V. In this
book, it is suggested that the adjustment shall not exceed 2% (Fig. 2.23).
The analysis of physical properties of 2D nMOSFET based on Sentaurus
Visual interface is as shown in Fig. 2.24.
2D-nMOSFET
L g = 200 nm
Vd = 3 V
Drain Current, Id (A)
Step = 200 nm
L g = 1000 nm
Fig. 2.23 Id–Vg curves of simulation of 2D nMOSFET. Some descriptions are added by
PowerPoint after snapshot from Inspect. It reveals that the 2D nMOSFET with Lg < 600 nm
suffers severe short-channel effect (SCE)
2.2 [Example 2.1] 2D n-Type MOSFET Id–Vg Characteristics Simulation 49
Fig. 2.24 Use Sentaurus Visual interface. Important device physical properties can be visualized
and analyzed
Figure 2.24 is the Sentaurus Visual interface. The Selection on the left is for
selecting the material to be displayed and its mesh. In addition, the physical
property to be inspected can be selected on the lower part of screen such as: energy
band diagram and carrier distribution of electrons and holes.
The electron concentration distribution, electric field distribution, electro-
static potential distribution, and energy band diagram along the channel
direction are as shown in Figs. 2.25, 2.26, 2.27, and 2.28 based on the conditions of
Lg = 1000 nm, Vd = 3 V, and Vg = 3 V.
Electron concentration
Lg =1000nm
Vd =3V
Vg =3V
S G D
Electric Field
Lg =1000nm
Vd =3V
Vg =3V
S G D
Fig. 2.26 Electric field distribution of the simulation of 2D n-type semiconductor device
Electric Field
Lg =1000nm
Vd =3V
Vg =3V
S G D
Fig. 2.27 Electrostatic potential distribution of the simulation of 2D n-type semiconductor device
2.3 [Example 2.2] 2D n-Type MOSFET with Id–Vd Characteristics Simulation 51
Band Diagram
Channel
Energy (eV)
Lg = 1000 nm
Vd = 3 V
Vg = 3 V
Fig. 2.28 Energy band diagram along the channel direction of the simulation of 2D n-type
semiconductor device
This Id–Vd example is very similar to Example 2.1, only has difference in electrodes
and its bias setting.
The following three main program code files are all based on Synopsys
Sentaurus TCAD 2014 version.
1. SDE—devise_dvs.cmd
This is the best example of 2D nMOSFET Id–Vd.
The line of code following; is the prompt character for program designer to
take note such that it will not be executed by the computer.
52 2 2D MOSFET Simulation
(position x3 0 0)
(position x4 y2 0) "Silicon" "Drain" )
(sdegeo:create-rectangle
(position x4 0 0)
(position x5 y2 0) "Silicon" "DrainC" )
;--- Body ---
(sdegeo:create-rectangle
(position 0 0 0)
(position x5 y1 0) "Silicon" "Body" )
;--- Gate oxide ---
(sdegeo:create-rectangle
(position x2 y2 0)
(position x3 y3 0) "SiO2" "Gateoxide" )
;------------------------ Contact -----------------------------;
;----- Gate -----
(sdegeo:define-contact-set "G"
4.0 (color:rgb 1.0 0.0 0.0 ) "##")
(sdegeo:define-2d-contact (find-edge-id (position (+ x2 (/ Lg 2)) y3 0)) "G")
;----- Source -----
(sdegeo:define-contact-set "S"
4.0 (color:rgb 1.0 0.0 0.0 ) "##")
(sdegeo:define-2d-contact (find-edge-id (position 10 tac 0)) "S")
;----- Drain -----
(sdegeo:define-contact-set "D"
4.0 (color:rgb 1.0 0.0 0.0 ) "##")
(sdegeo:define-2d-contact (find-edge-id (position (+ 50 Lg 35) tac 0)) "D")
;----- Substrate -----
(sdegeo:define-contact-set "substrate"
4.0 (color:rgb 1.0 0.0 0.0 ) "##")
(sdegeo:define-2d-contact (find-edge-id (position (+ x2 (/ Lg 2)) (- Body) 0))
"substrate")
;---------------- Doping ----------------;
;--- Channel ---
(sdedr:define-constant-profile "dopedC" "BoronActiveConcentration" C_Doping )
(sdedr:define-constant-profile-region "RegionC" "dopedC" "Channel" )
54 2 2D MOSFET Simulation
2 SDVICE—dessis_des.cmd
The line of code following # and * are the prompt character for program
designer to take note such that it will not be executed by the computer.
Electrode{
{name="D" voltage=0.0}
{name="S" voltage=0.0}
{name="G" voltage=0.0 WorkFunction=@WK@}
}
File{
Grid="@tdr@"
Plot="@tdrdat@"
Current="@plot@"
Output="@log@"
parameter="@parameter@"
}
Physics{
Mobility( DopingDep HighFieldSaturation Enormal )
EffectiveIntrinsicDensity( OldSlotboom )
Recombination( SRH(DopingDep) )
eQuantumPotential
}
Math{
-CheckUndefinedModels
Number_Of_Threads=4
Extrapolate
Derivatives
* Avalderivatives
RelErrControl
Digits=5
56 2 2D MOSFET Simulation
ErRef(electron)=1.e10
ErRef(hole)=1.e10
Notdamped=50
Iterations=20
Directcurrent
Method=ParDiSo
Parallel= 2
*-VoronoiFaceBoxMethod
NaturalBoxMethod
}
Plot{
eDensity hDensity
eCurrent hCurrent
TotalCurrent/Vector eCurrent/Vector hCurrent/Vector
eMobility hMobility
eVelocity hVelocity
eEnormal hEnormal
ElectricField/Vector Potential SpaceCharge
eQuasiFermi hQuasiFermi
Potential Doping SpaceCharge
SRH Auger
AvalancheGeneration
DonorConcentration AcceptorConcentration
Doping
eGradQuasiFermi/Vector hGradQuasiFermi/Vector
eEparallel hEparalllel
BandGap
BandGapNarrowing
Affinity
ConductionBand ValenceBand
eQuantumPotential
}
Solve {
Coupled ( Iterations= 150){ Poisson eQuantumPotential }
Coupled { Poisson eQuantumPotential Electron Hole }
2.3 [Example 2.2] 2D n-Type MOSFET with Id–Vd Characteristics Simulation 57
Quasistationary(
InitialStep= 1e-3 Increment= 1.2
MinStep= 1e-12 MaxStep= 0.02
Goal { Name= "G" Voltage=@Vg@ }
){ Coupled { Poisson eQuantumPotential Electron Hole } }
Quasistationary(
InitialStep= 1e-3 Increment= 1.2
MinStep= 1e-12 MaxStep= 0.95
Goal { Name= "D" Voltage=@Vd@ }
DoZero
){ Coupled { Poisson eQuantumPotential Electron Hole } }
}
*------------------- END -------------------*
3. INSPECT—inspect_inc.cmd
The line of code following # is the prompt character for program designer to take
note such that it will not be executed by the computer.
#-----------------------------------------------------------------------#
# Script file designed to compute : #
# * The threshold voltage : VT #
# * The transconductance : gm #
#-----------------------------------------------------------------------#
if { ! [catch {open n@previous@_ins.log w} log_file] } {
set fileId stdout
}
puts $log_file " "
puts $log_file " ------------------------------------ "
puts $log_file " Values of the extracted Parameters : "
puts $log_file " ------------------------------------ "
puts $log_file " "
58 2 2D MOSFET Simulation
2D-nMOSFET, Lg = 1000 nm
Vg = 3 V
Drain Current, Id (A)
Vg = 2 V
Vg = 1 V
The following three main program code files are all based on Synopsys
Sentaurus TCAD 2014 version.
This 2D pMOSFET Id–Vg simulation example is very similar to Example 2.1
nMOSFET Id–Vg, only have difference in doping and electrodes bias setting. This is
the standard example of 2D pMOSFET Id–Vg example.
1. SDE – devise_dvs.cmd
This is the best example of 2D p MOSFET.
The line of code following; is the prompt character for program designer to
take note such that it will not be executed by the computer.
2.4 [Example 2.3] 2D p-Type MOSFET with Id–Vg Characteristics Simulation 61
(position x3 0 0)
(position x4 y2 0) "Silicon" "Drain" )
(sdegeo:create-rectangle
(position x4 0 0)
(position x5 y2 0) "Silicon" "DrainC" )
;--- Body ---
(sdegeo:create-rectangle
(position 0 0 0)
(position x5 y1 0) "Silicon" "Body" )
;--- Gate oxide ---
(sdegeo:create-rectangle
(position x2 y2 0)
(position x3 y3 0) "SiO2" "Gateoxide" )
;----------------------- Contact ------------------------;
;----- Gate -----
(sdegeo:define-contact-set "G"
4.0 (color:rgb 1.0 0.0 0.0 ) "##")
(sdegeo:define-2d-contact (find-edge-id (position (+ x2 (/ Lg 2)) y3 0)) "G")
;----- Source -----
(sdegeo:define-contact-set "S"
4.0 (color:rgb 1.0 0.0 0.0 ) "##")
(sdegeo:define-2d-contact (find-edge-id (position 10 tac 0)) "S")
;----- Drain -----
(sdegeo:define-contact-set "D"
4.0 (color:rgb 1.0 0.0 0.0 ) "##")
(sdegeo:define-2d-contact (find-edge-id (position (+ 50 Lg 35) tac 0)) "D")
;----- Substrate -----
(sdegeo:define-contact-set "substrate"
4.0 (color:rgb 1.0 0.0 0.0 ) "##")
(sdegeo:define-2d-contact (find-edge-id (position (+ x2 (/ Lg 2)) (- Body) 0))
"substrate")
;---------------------- Doping -------------------------;
;--- Channel ---
(sdedr:define-constant-profile "dopedC" "PhosphorusActiveConcentration"
C_Doping )
2.4 [Example 2.3] 2D p-Type MOSFET with Id–Vg Characteristics Simulation 63
(sdegeo:set-default-boolean "XX")
(sdeio:read-tdr-bnd tdrin)
(entity:scale (get-body-list) sf)
(sdeio:save-tdr-bnd (get-body-list) tdrout)
)
)
(sde:scale-tdr-bnd "n@node@_nm.tdr" nm "n@node@_bnd.tdr")
;---------------- END -----------------;
2. SDVICE – dessis_des.cmd
The line of code following * and # are the prompt characters for program
designer to take note such that it will not be executed by the computer.
Electrode{
{name="D" voltage=0.0}
{name="S" voltage=0.0}
{name="G" voltage=0.0 WorkFunction=@WK@}
}
File{
Grid="@tdr@"
Plot="@tdrdat@"
Current="@plot@"
Output="@log@"
parameter="@parameter@"
}
Physics{
Mobility( DopingDep HighFieldSaturation Enormal )
EffectiveIntrinsicDensity( OldSlotboom )
2.4 [Example 2.3] 2D p-Type MOSFET with Id–Vg Characteristics Simulation 65
Recombination( SRH(DopingDep) )
hQuantumPotential
}
Math{
-CheckUndefinedModels
Number_Of_Threads=4
Extrapolate
Derivatives
* Avalderivatives
RelErrControl
Digits=5
ErRef(electron)=1.e10
ErRef(hole)=1.e10
Notdamped=50
Iterations=20
Directcurrent
Method=ParDiSo
Parallel= 2
*-VoronoiFaceBoxMethod
NaturalBoxMethod
}
Plot{
eDensity hDensity
eCurrent hCurrent
TotalCurrent/Vector eCurrent/Vector hCurrent/Vector
eMobility hMobility
eVelocity hVelocity
eEnormal hEnormal
ElectricField/Vector Potential SpaceCharge
eQuasiFermi hQuasiFermi
Potential Doping SpaceCharge
SRH Auger
AvalancheGeneration
66 2 2D MOSFET Simulation
DonorConcentration AcceptorConcentration
Doping
eGradQuasiFermi/Vector hGradQuasiFermi/Vector
eEparallel hEparalllel
BandGap
BandGapNarrowing
Affinity
ConductionBand ValenceBand
hQuantumPotential
}
Solve {
Coupled ( Iterations= 150){ Poisson hQuantumPotential }
Coupled { Poisson hQuantumPotential Electron Hole }
Quasistationary(
InitialStep= 1e-3 Increment= 1.2
MinStep= 1e-12 MaxStep= 0.95
Goal { Name= "D" Voltage=@Vd@ }
){ Coupled { Poisson hQuantumPotential Electron Hole } }
Quasistationary(
InitialStep= 1e-3 Increment= 1.2
MinStep= 1e-12 MaxStep= 0.02
Goal { Name= "G" Voltage=@Vg@ }
DoZero
){ Coupled { Poisson hQuantumPotential Electron Hole } }
}
*--------------- END ---------------*
2.4 [Example 2.3] 2D p-Type MOSFET with Id–Vg Characteristics Simulation 67
3. INSPECT – inspect_inc.cmd
The line of code following # is the prompt character for program designer to
take note such that it will not be executed by the computer.
#------------------------------------------------------------------------#
# Script file designed to compute : #
# * The threshold voltage : VT
#
# * The transconductance : gm #
#------------------------------------------------------------------------#
if { ! [catch {open n@previous@_ins.log w} log_file] } {
set fileId stdout
}
puts $log_file " "
puts $log_file " ------------------------------------ "
puts $log_file " Values of the extracted Parameters : "
puts $log_file " ------------------------------------ "
puts $log_file " "
puts $log_file " "
set DATE [ exec date ]
set WORK [ exec pwd ]
puts $log_file " Date : $DATE "
puts $log_file " Directory : $WORK "
puts $log_file " "
puts $log_file " "
# #
# idvgs=y(x) ; vgsvgs=x(x) ; #
#set out_file n@previous@_des
proj_load "${out_file}.plt"
# ---------------------------------------------------------------------- #
# I) VT = Xintercept(maxslope(ID[VGS])) or VT = VGS( IDS= 100nA/um ) #
# ---------------------------------------------------------------------- #
cv_create idvgs "${out_file} G OuterVoltage" "${out_file} S TotalCurrent"
cv_create vdsvgs "${out_file} G OuterVoltage" "${out_file} S OuterVoltage"
#....................................................................... #
# 1) VT extracted as the intersection point with the X axis at the point #
# where the id(vgs) slope reaches its maxmimum : #
#....................................................................... #
set VT1 [ f_VT1 idvgs ]
68 2 2D MOSFET Simulation
#................................................................ #
# 2) Printing of the whole set of extracted values (std output) : #
#................................................................ #
puts $log_file "Threshold voltage VT1 = $VT1 Volts"
puts $log_file " "
#...................................................................... #
# 3) Initialization and display of curves on the main Inspect screen : #
cv_display idvgs
cv_lineStyle idvgs solid
cv_lineColor idvgs red
# ---------------------------------------------------------------------- #
# II) gm = maxslope((ID[VGS]) #
# ---------------------------------------------------------------------- #
set gm [ f_gm idvgs ]
puts $log_file " "
puts $log_file "Transconductance gm = $gm A/V"
puts $log_file " "
set ioff [ cv_compute "vecmin(<idvgs>)" A A A A ]
puts $log_file " "
puts $log_file "Current ioff = $ioff A"
puts $log_file " "
set isat [ cv_compute "vecmax(<idvgs>)" A A A A ]
puts $log_file " "
puts $log_file "Current isat = $isat A"
puts $log_file " "
set rout [ cv_compute "Rout(<idvgs>)" A A A A ]
puts $log_file " "
puts $log_file "Resistant rout = $rout A"
puts $log_file " "
cv_createWithFormula logcurve "log10(<idvgs>)" A A A A
cv_createWithFormula difflog "(-1)*diff(<logcurve>)" A A A A
set sslop [ cv_compute "1/vecmax(<difflog>)" A A A A ]
puts $log_file " "
puts $log_file "sub solp = $sslop A/V"
puts $log_file " "
2.4 [Example 2.3] 2D p-Type MOSFET with Id–Vg Characteristics Simulation 69
2D-pMOSFET, Lg = 1000 nm
Vd = -3 V
Drain Current, Id (A)
Vd = -0.1 V
S.S. = 89 mV/dec @ Vd = -3 V
1. SDE – devise_dvs.cmd
The line of code following ; is the prompt character for program designer to
take note such that it will not be executed by the computer.
(position x3 0 0)
(position x4 y2 0) "Silicon" "Drain" )
(sdegeo:create-rectangle
(position x4 0 0)
(position x5 y2 0) "Silicon" "DrainC" )
;--- Body ---
(sdegeo:create-rectangle
(position 0 0 0)
(position x5 y1 0) "Silicon" "Body" )
;--- Gate oxide ---
(sdegeo:create-rectangle
(position x2 y2 0)
(position x3 y3 0) "SiO2" "Gateoxide" )
;------------------------- Contact --------------------------;
;----- Gate -----
(sdegeo:define-contact-set "G"
4.0 (color:rgb 1.0 0.0 0.0 ) "##")
(sdegeo:define-2d-contact (find-edge-id (position (+ x2 (/ Lg 2)) y3 0)) "G")
;----- Source -----
(sdegeo:define-contact-set "S"
4.0 (color:rgb 1.0 0.0 0.0 ) "##")
(sdegeo:define-2d-contact (find-edge-id (position 10 tac 0)) "S")
;----- Drain -----
(sdegeo:define-contact-set "D"
4.0 (color:rgb 1.0 0.0 0.0 ) "##")
(sdegeo:define-2d-contact (find-edge-id (position (+ 50 Lg 35) tac 0)) "D")
;----- Substrate -----
(sdegeo:define-contact-set "substrate"
4.0 (color:rgb 1.0 0.0 0.0 ) "##")
(sdegeo:define-2d-contact (find-edge-id (position (+ x2 (/ Lg 2)) (- Body) 0))
"substrate")
;----------------------------- Doping ------------------------------;
;--- Channel ---
(sdedr:define-constant-profile "dopedC" "PhosphorusActiveConcentration"
C_Doping )
72 2 2D MOSFET Simulation
(sdegeo:set-default-boolean "XX")
(sdeio:read-tdr-bnd tdrin)
(entity:scale (get-body-list) sf)
(sdeio:save-tdr-bnd (get-body-list) tdrout)
)
)
(sde:scale-tdr-bnd "n@node@_nm.tdr" nm "n@node@_bnd.tdr")
;------------------------------ END ---------------------------------;
2. SDVICE – dessis_des.cmd
The line of code following # and * are the prompt characters for program
designer to take note such that it will not be executed by the computer.
Electrode{
{name="D" voltage=0.0}
{name="S" voltage=0.0}
{name="G" voltage=0.0 WorkFunction=@WK@}
}
File{
Grid="@tdr@"
Plot="@tdrdat@"
Current="@plot@"
Output="@log@"
parameter="@parameter@"
}
Physics{
Mobility( DopingDep HighFieldSaturation Enormal )
74 2 2D MOSFET Simulation
EffectiveIntrinsicDensity( OldSlotboom )
Recombination( SRH(DopingDep) )
hQuantumPotential
}
Math{
-CheckUndefinedModels
Number_Of_Threads=4
Extrapolate
Derivatives
* Avalderivatives
RelErrControl
Digits=5
ErRef(electron)=1.e10
ErRef(hole)=1.e10
Notdamped=50
Iterations=20
Directcurrent
Method=ParDiSo
Parallel= 2
*-VoronoiFaceBoxMethod
NaturalBoxMethod
}
Plot{
eDensity hDensity
eCurrent hCurrent
TotalCurrent/Vector eCurrent/Vector hCurrent/Vector
eMobility hMobility
eVelocity hVelocity
eEnormal hEnormal
ElectricField/Vector Potential SpaceCharge
eQuasiFermi hQuasiFermi
Potential Doping SpaceCharge
SRH Auger
2.5 [Example 2.4] 2D p-Type MOSFET with Id–Vg Characteristics Simulation 75
AvalancheGeneration
DonorConcentration AcceptorConcentration
Doping
eGradQuasiFermi/Vector hGradQuasiFermi/Vector
eEparallel hEparalllel
BandGap
BandGapNarrowing
Affinity
ConductionBand ValenceBand
hQuantumPotential
}
Solve {
Coupled ( Iterations= 150){ Poisson hQuantumPotential }
Coupled { Poisson hQuantumPotential Electron Hole }
Quasistationary(
InitialStep= 1e-3 Increment= 1.2
MinStep= 1e-12 MaxStep= 0.02
Goal { Name= "G" Voltage=@Vg@ }
){ Coupled { Poisson hQuantumPotential Electron Hole } }
Quasistationary(
InitialStep= 1e-3 Increment= 1.2
MinStep= 1e-12 MaxStep= 0.95
Goal { Name= "D" Voltage=@Vd@ }
DoZero
){ Coupled { Poisson hQuantumPotential Electron Hole } }
}
*------------------------- END --------------------------*
76 2 2D MOSFET Simulation
3. INSPECT – inspect_inc.cmd
The line of code following # is the prompt character for program designer to take
note such that it will not be executed by the computer.
#------------------------------------------------------------------------#
# Script file designed to compute : #
# * The threshold voltage : VT
# * The transconductance : gm #
#------------------------------------------------------------------------#
if { ! [catch {open n@previous@_ins.log w} log_file] } {
set fileId stdout
}
puts $log_file " "
puts $log_file " ------------------------------------ "
puts $log_file " Values of the extracted Parameters : "
puts $log_file " ------------------------------------ "
puts $log_file " "
puts $log_file " "
set DATE [ exec date ]
set WORK [ exec pwd ]
puts $log_file " Date : $DATE "
puts $log_file " Directory : $WORK "
puts $log_file " "
puts $log_file " "
# #
# idvgs=y(x) ; vgsvgs=x(x) ; #
set out_file n@previous@_des
2.5 [Example 2.4] 2D p-Type MOSFET with Id–Vg Characteristics Simulation 77
proj_load "${out_file}.plt"
# I) VT = Xintercept(maxslope(ID[VGS])) or VT = VGS( IDS= 0.1 ua/um ) #
# ---------------------------------------------------------------------- #
cv_create idvgs "${out_file} G OuterVoltage" "${out_file} S TotalCurrent"
cv_create vdsvgs "${out_file} G OuterVoltage" "${out_file} S OuterVoltage"
#....................................................................... #
# 1) VT extracted as the intersection point with the X axis at the point #
# where the id(vgs) slope reaches its maxmimum : #
#....................................................................... #
set VT1 [ f_VT1 idvgs ]
#................................................................ #
# 2) Printing of the whole set of extracted values (std output) : #
#................................................................ #
puts $log_file "Threshold voltage VT1 = $VT1 Volts"
puts $log_file " "
#...................................................................... #
# 3) Initialization and display of curves on the main Inspect screen : #
# ..................................................................... #
cv_display idvgs
cv_lineStyle idvgs solid
cv_lineColor idvgs red
# ---------------------------------------------------------------------- #
# II) gm = maxslope((ID[VGS]) #
# ---------------------------------------------------------------------- #
set gm [ f_gm idvgs ]
puts $log_file " "
puts $log_file "Transconductance gm = $gm A/V"
puts $log_file " "
set ioff [ cv_compute "vecmin(<idvgs>)" A A A A ]
78 2 2D MOSFET Simulation
2D-pMOSFET, Lg = 1000 nm
Vg = -3 V
Drain Current, Id (A)
Vg = -2 V
Vg = -1 V
Gate
Source Drain
Channel
LDD LDD
Body
the spatial charge region will be reduced, thus minimizing the breakdown effect. As
for the peak value at drain junction, the electric field is the function of semicon-
ductor doping and the function of curvature of n+ drain region. In the LDD
structure, the electric field of oxide-semiconductor junction is lower than the tra-
ditional structure. Among traditional devices, electric fields usually peak at the
metallurgical junction, and it will be quickly reduced to zero at the drain. This is
because the electric field cannot exist in the highly conductive n+ region. On the
other hand, the electric field in LDD device will be extended across the n region
before being reduced to zero, and this effect will minimize the breakdown effect and
hot carrier effect.
There are two disadvantages of LDD device. For one, the fabrication complexity
is increased. For the other, the drain resistance is increased. Nonetheless, this extra
process step can indeed fabricate the device with significantly improved perfor-
mance. The cross section of LDD device is as shown in Fig. 2.32, in which the
source terminal is changed to the lightly doped n region, which will lead to improve
the device operating performance while reducing the process complexity. The series
resistance will lead to increased device power consumption, so this factor must be
taken into consideration for high-power device.
2.6 [Example 2.5] 2D n-Type MOSFET with LDD (Lightly Doped Drain) Simulation 81
1. SDE – devise_dvs.cmd
The line of code following ; is the prompt character for program designer to
take note such that it will not be executed by the computer;
;------------------------ parameter -----------------------------;
(define Lg @Lg@)
(define tox @tox@)
(define tac 100)
(define Body 400)
(define LSDC 25)
(define LSD 25)
(define C_Doping 1e16)
(define DC_Doping 1e20)
(define D_Doping 1e18)
(define S_Doping 1e18)
(define SC_Doping 1e20)
;(define B_Doping 1e15)
(define B_Doping 1e16)
(define nm 1e-3)
(define x1 LSDC)
(define x2 (+ x1 LSD))
(define x3 (+ x2 Lg))
(define x4 (+ x3 LSD))
(define x5 (+ x4 LSDC))
(define y1 (- Body))
(define y2 tac)
(define y3 (+ tac tox))
;-------------------------- Structure -------------------------;
"ABA"
;--- source ---
(sdegeo:create-rectangle
(position 0 0 0)
(position x1 y2 0) "Silicon" "SourceC" )
(sdegeo:create-rectangle
(position x1 0 0)
(position x2 y2 0) "Silicon" "Source" )
;--- Channel ---
(sdegeo:create-rectangle
(position x2 0 0)
82 2 2D MOSFET Simulation
2. SDVICE – dessis_des.cmd
The line of code following # and * are the prompt characters for program
designer to take note such that it will not be executed by the computer.
Electrode{
{name="D" voltage=0.0}
{name="S" voltage=0.0}
{name="G" voltage=0.0 WorkFunction=@WK@}
}
File{
Grid="@tdr@"
Plot="@tdrdat@"
Current="@plot@"
Output="@log@"
parameter="@parameter@"
}
Physics{
Mobility( DopingDep HighFieldSaturation Enormal )
EffectiveIntrinsicDensity( OldSlotboom )
Recombination( SRH(DopingDep) )
eQuantumPotential
}
Math{
-CheckUndefinedModels
2.6 [Example 2.5] 2D n-Type MOSFET with LDD (Lightly Doped Drain) Simulation 85
Number_Of_Threads=4
Extrapolate
Derivatives
* Avalderivatives
RelErrControl
Digits=5
ErRef(electron)=1.e10
ErRef(hole)=1.e10
Notdamped=50
Iterations=20
Directcurrent
Method=ParDiSo
Parallel= 2
*-VoronoiFaceBoxMethod
NaturalBoxMethod
}
Plot{
eDensity hDensity
eCurrent hCurrent
TotalCurrent/Vector eCurrent/Vector hCurrent/Vector
eMobility hMobility
eVelocity hVelocity
eEnormal hEnormal
ElectricField/Vector Potential SpaceCharge
eQuasiFermi hQuasiFermi
Potential Doping SpaceCharge
SRH Auger
AvalancheGeneration
DonorConcentration AcceptorConcentration
Doping
eGradQuasiFermi/Vector hGradQuasiFermi/Vector
eEparallel hEparalllel
BandGap
BandGapNarrowing
86 2 2D MOSFET Simulation
Affinity
ConductionBand ValenceBand
eQuantumPotential
}
Solve {
Coupled ( Iterations= 150){ Poisson eQuantumPotential }
Coupled { Poisson eQuantumPotential Electron Hole }
Quasistationary(
InitialStep= 1e-3 Increment= 1.2
MinStep= 1e-12 MaxStep= 0.95
Goal { Name= "D" Voltage=@Vd@ }
){ Coupled { Poisson eQuantumPotential Electron Hole } }
Quasistationary(
InitialStep= 1e-3 Increment= 1.2
MinStep= 1e-12 MaxStep= 0.02
Goal { Name= "G" Voltage=@Vg@ }
DoZero
){ Coupled { Poisson eQuantumPotential Electron Hole } }
}
*----------------------------------------- END ------------------------------------------*
3. INSPECT – inspect_inc.cmd
The line of code following # is the prompt character for program designer to take
note such that it will not be executed by the computer.
2.6 [Example 2.5] 2D n-Type MOSFET with LDD (Lightly Doped Drain) Simulation 87
#------------------------------------------------------------------------#
# Script file designed to compute : #
# * The threshold voltage : VT# #
# * The transconductance : gm #
#------------------------------------------------------------------------#
if { ! [catch {open n@previous@_ins.log w} log_file] } {
set fileId stdout
}
puts $log_file " "
puts $log_file " ------------------------------------ "
puts $log_file " Values of the extracted Parameters : "
puts $log_file " ------------------------------------ "
puts $log_file " "
puts $log_file " "
set DATE [ exec date ]
set WORK [ exec pwd ]
puts $log_file " Date : $DATE "
puts $log_file " Directory : $WORK "
puts $log_file " "
puts $log_file " "
# #
# idvgs=y(x) ; vgsvgs=x(x) ; #
set out_file n@previous@_des
proj_load "${out_file}.plt"
# ---------------------------------------------------------------------- #
# I) VT = Xintercept(maxslope(ID[VGS])) or VT = VGS( IDS= 0.1 ua/um ) #
# ---------------------------------------------------------------------- #
cv_create idvgs "${out_file} G OuterVoltage" "${out_file} D TotalCurrent"
cv_create vdsvgs "${out_file} G OuterVoltage" "${out_file} D OuterVoltage"
#....................................................................... #
# 1) VT extracted as the intersection point with the X axis at the point #
# where the id(vgs) slope reaches its maxmimum : #
#....................................................................... #
set VT1 [ f_VT1 idvgs ]
#................................................................ #
88 2 2D MOSFET Simulation
The electric property is as shown in Fig. 2.33, which shows that when LDD
(lightly doped drain) is added into the original device, the leakage current (Ioff) is
reduced and the sub-threshold slope (SS) is significantly improved as compared to
Fig. 2.23. Using LDD for reducing short-channel effect, the 2D n-type MOSFET
still has good performance at Lg = 400 nm.
Lg=200nm
Vd =3V
Step=200nm
Id(A)
2D LDD nMOSFET
Vg (V)
2.7 Summary
References
1. S.M. Sze, K.K. Ng, Physics of Semiconductor Devices, 3rd ed (Wiley, New York, 2007)
2. TCAD Sentaurus Device, Synopsys SDevice Ver. J-2014.09 (Synopsys, Inc., Mountain View,
CA, USA, 2014)
3. C.C. Hu, Modern Semiconductor Devices for Integrated Circuits (Pearson Education, Inc.,
2010)
Chapter 3
3D FinFET with Lg = 15 nm
and Lg = 10 nm Simulation
In 1965, Gordon Moore proposed the rule that the number of devices on the wafer
would be doubled every 18–24 months. This “Moore’s law” describes the con-
tinuous and rapid trend of scaling. Every reduction of feature size will be called a
technology generation or technology node. The technology nodes include 0.18,
0.13, 90, 65, and 45 lm. These numbers represent the minimum feature size. For
every new technology node, all circuit layout properties (such as SRAM cell and
CPU) will have their size reduced to 70%. From the historical perspective, new
technology node will be generated once every 2 or 3 years. The advantage of new
technology node is that the circuit size can be reduced in half of original feature
length (70%) and width means 50% reduction of area rate (0.7 0.7 = 0.49).
For every new technology node, there can be twice as many circuits on one wafer
such that the cost of every circuit will be greatly reduced, thus reducing the cost of
IC. Moreover, by scaling, the device performances will also enhance such as IC
operation speed increasing and power reduction [1].
In addition to gate length (Lg) and width W scaling down, many other param-
eters will be reduced along with the scaling rule. For example, the effective
thickness of gate oxide (EOT) of MOSFET and power supply voltage (Vdd) will be
also reduced to increase the transistor current Ion density (Ion/W) and decrease the
circuit operation power. Smaller transistor and shorter internal connection will
result in smaller capacitance, and these changes will all reduce the circuit delay
time. Historically, the IC speed has grown by 30% at every new technology node.
Higher speed will result in innovative IC applications, such as the higher CPU
speed, higher DRAM and flash memory density, and higher broadband data
transmission via RF circuit for cell phone. A good reviewing report is presented in
Intel 22-nm FinFET study [2].
MPU/SoC Metalx V2 pitch (nm) 28.0 18.0 12.0 10.0 6.0 6.0 6.0
MPU/SoC Metal0/1 Vi pitch (nm) 28.0 18.0 12.0 10.0 6.0 6.0 6.0
Lg: physical gate length for HP logic 24 18 14 10 10 10 10
(nm)
Lg: physical gate length for LP logic 26 20 16 12 12 12 12
(nm)
FinFET Fin width (nm) 8.0 6.0 6.0 NA N/A N/A N/A
FinFET Fin height (nm) 42.0 42.0 42.0 NA N/A N/A N/A
Device effective width (nm) 92.0 90.0 56.5 56.5 56.5 56.5 56.5
Device lateral half pitch (nm) 21.0 18.0 12.0 10.0 6.0 6.0 6.0
Device width or diameter (nm) 8.0 6.0 6.0 6.0 5.0 5.0 5.0
Device physical and electrical specs
Power supply voltage Vdd (V) 0.80 0.75 0.70 0.65 0.55 0.45 0.40
Subthreshold slope (mV/dec) 75 70 68 65 40 25 25
Inversion layer thickness (nm) 1.10 1.00 0.90 0.85 0.80 0.80 0.80
Vt,sat (mV) at Ioff = 100 nA/lm 129 129 133 136 84 52 52
(HP logic)
Vt,sat (mV) at Ioff = 100 pA/lm 351 336 333 326 201 125 125
(LP logic)
(continued)
93
Table 3.1 (continued)
94
Ioff 1E-11
Vg
Vth= 0.15 0.22 0.3
Fig. 3.1 Id–Vg curve of N14/N16 HP, SP, LSTP nFinFET important parameters
The circuit speed will be increased along with increasing Ion, thus requiring a
smaller threshold voltage. However, the main current of MOSFET in the off state is
Ioff, and the Id is the value of Ioff measured with Vgs = 0 and Vds = Vdd as shown in
96 3 3D FinFET with Lg = 15 nm and Lg = 10 nm Simulation
Lg=12nm
Fwt=4nm
D
Fh =40nm G
S
Fwb=6nm
STI SiO2
Eq. (3.1). The minimization of Ioff is the most important task for minimizing the
power consumption by circuit in standby mode [1, 3].
Practically, Vth is usually defined as the Vgs when Ids = 100 nA (W/L) as
shown in Eq. (3.1). MOSFET off state current (Ioff) is defined by Eq. (3.2). The
Eq. (3.1) can be substituted into obtain the equation of relationship between Ioff and
Vth, which is as shown as Eq. (3.3). Wherein, the simplification of Eq. (3.1) is
another reason for Vth definition, meaning that the function exp(qVgs/kT) will be
changed by 10 whenever Vgs is changed by 60 mV under room temperature, such
that exp(qVgs/ηkT) will be changed by 10 times at every η 60 mV. For example,
if η = 1.5 and Vgs < Vth under room temperature, it is indicated in Eq. (3.1) that Ids
will be reduced 10 times along with every decline of Vgs by 90 mV. η 60 mV is
called subthreshold swing represented by the symbol SS as shown in Eq. (3.4) [1].
W qðVgs VtÞ=gkT W
Ids ðnAÞ ¼ 100 e ¼ 100 10ðVgs VtÞ=SS ð3:1Þ
L L
Vds ¼ 0; Ids ¼ Ioff ð3:2Þ
W qVt=gkT W
Ioff ðnAÞ ¼ 100 e ¼ 100 10Vt=SS ð3:3Þ
L L
SSðmV=decadeÞ ¼ g 60 mV ð3:4Þ
From (3.2)
kT
SS ¼ g ln 10 ¼ g ð26 mV 2:3Þ ¼ g 60 mV at 300 K ð3:6Þ
q
Cdep
SS ¼ 60 mV g *g ¼ 1 þ [1 )usually SS [ 60 mV ð3:7Þ
Coxe
As for the assigned W and L, there are two approaches for minimizing Ioff. The
first approach is to select higher Vth, yet this is not the optimal solution because
higher Vth will lead to Ion reduction, thus lowering the circuit speed. Another better
approach is to reduce subthreshold swing (SS) by increasing Coxe to reduce η,
which means the thinner gate oxide thickness (Tox) is to be used. Using FinFET
with high-k dielectric materials and metal gate can approach ideal value
SS = 60 mV/decade. There is yet another approach for reducing Ioff by SS
reduction, which is allowing the transistor to be operated in low temperature.
However, the low-temperature operation may lead to significant increase of cost.
Vth will be reduced along with the scaling of L as shown in Fig. 3.3. With
significant reduction of Vth, Ioff will become rather high, thus worsening the channel
leakage current. For increasing Vth, the doping concentration of the body (Nb) of
short-channel device will be higher than the long-channel device. The energy band
diagram is shown for the long- and short-channel semiconductor–insulator junction
in Fig. 3.3a, c with Vgs = 0. Figure 3.3b shows the case at Vgs = Vth. In the case of
(b), Ec in the channel is pulled lower than in the case of (a), and therefore is closer
to the Ec of source. In this case, electrons can flow from N+ source through the
channel to the drain. Figure 3.3c shows the case of short-channel device at Vgs = 0.
If the channel is short enough, Ec will not be able to reach the same peak value as in
(a). As the results, Vth value is lower in short-channel device than that of the
long-channel device. The decreasing of Vth value can be explained as Vth roll-off.
Therefore, Vth must be set in reasonable range for different gate lengths. Currently,
for FinFET N16/N14 node, Vtn is 0.15–0.35 V and Vtp is −0.15 to
−0.35 V (Figs. 3.4 and 3.5).
98 3 3D FinFET with Lg = 15 nm and Lg = 10 nm Simulation
Vgs = 0 V Vgs = 0 V
Ec Ef
N+ Source
N+ Drain
(b) (d)
Vgs = Vt-long Vgs = Vt-short
~0.2 V
Ef
Fig. 3.3 a–d Energy band diagram of source to drain with Vgs = 0 V and Vgs = Vth, a–b long
channel; c–d short channel
P-Sub
With device size scaling, the thickness of gate oxide should be reduced along with
the scaling of channel size. However, if the gate oxide is too thin, it will induce
severe gate tunneling current, which will increase the device off current (Ioff), thus
leading to increased standby power consumption of the portable 3C products.
Therefore, the severe gate leakage current will work against effective scaling down
of device size.
Nowadays, the applications of high-k dielectric materials in semiconductor
industry have been used [1]. With the requirements of reducing device dimension
3.3 Design Considerations of High-k Dielectric Materials … 99
S Cd D
Cg
Leakage path
7
ZrO2 HfO 2
6
4
0 5 10 15 20 25 30
K
and gate leakage current, many gate dielectric materials with high-k value have
been proposed to replace the traditional SiO2 gate dielectric layer, including Al2O3,
HfO2, ZrO2, and La2O3. Currently, in N14/N16 FinFET, the mainstream high-
k dielectric material is HfO2 with relative dielectric constant k of 24. The rela-
tionship between energy gaps (Eg) and relative dielectric constants (k) of various
materials is as shown in Fig. 3.6.
With the same equivalent oxide thickness (EOT), the use of high-k material can
reduce the gate leakage current, because the high-k material is higher than the
traditional SiO2 dielectric constant of 3.9, thus leading to higher relative capaci-
tance in identical thickness. The higher gate capacitance increases higher driving
current as shown in Eq. (3.8). The description and example are as shown below:
kox e0 khk e0
Cgate ¼ ¼ ðF=cm2 Þ ð3:8Þ
tox thk
kox
EOT ¼ thk ¼ tox ð3:9Þ
khk
For example; For HfO2 ; thk ¼ 6 nm; khk ¼ 24; kox ¼ 3:9; )EOT ¼ 1 nm
For N14/N16 FinFET; HfO2 : thk ¼ 3 nm; then EOT ¼ 0:5 nm
100 3 3D FinFET with Lg = 15 nm and Lg = 10 nm Simulation
In addition to the impact of high-k material, the work functions (WF) of different
metal gates being used will also affect Vth [4]; the relationship between metal work
function and Vth can be substituted into Vth Eq. (3.10) via Eq. (3.11) in order to
obtain the relationship equation as Vtn of nFinFET Eq. (3.12) and Vtp of pFinFET
(3.13).
Qf
Vtn ¼ Vfb þ /S þ Vhk ¼ wmN wS þ /S þ Vhk
Chk
Qf
¼ wmN wS þ /S þ Vhk ð3:12Þ
Chk
Qf
Vtp ¼ Vfb /S Vhk ¼ wmP wS /S Vhk
C hk
Qf
¼ wmP wS /S Vhk ð3:13Þ
Chk
Vacuum level
4.05 eV
4.61 eV
5.17 eV
N+ poly-Si 4.05 eV
Ec N-type Metal ψmN < 4.2 eV
Fig. 3.7 Work function values of various metal frequently used in FinFET
The evolution of gate started from the initial planar gate to double gate and then
advanced to current tri-gate and gate-all-around (GAA) as shown in Fig. 3.8. The
more gate numbers covering the channel lead to better control capability [5]. The
electrical line of electric field established by the multigate can be more focused and
penetrating deeper in the channel to prevent electric field of drain. This effect can
reduce the short-channel effect. The 3D electric field distribution within the tran-
sistor channel is as shown in Fig. 3.9.
From Poisson’s equation, it is shown that the solution form of electrical potential
of drain is as follows: uðxÞ ¼ u0 expð kx1 Þ, where k is defined as the natural
length.
A small k value will result in rapidly reducing electrical potential of drain
such that it will have lesser impact on the channel. This effect will help gate to
dominate the transistor switching. The relationship equations of natural lengths
under different gate structures are as shown in Table 3.3. The multigate structure
like FinFET and gate-all-around FET (GAA FET) have smaller k and leading to
reduction of short-channel effect.
The smaller natural length will lead to better device performance. In general, the
reasonable choice is Lg > 5–10 times of k. We use the aforementioned equations
derived from the square as the equivalent rectangular FinFET for approximation
calculation. For example, in Fig. 3.10, Tsi Tsi = Fw Fh; the 10 nm 10 nm
square nanowire FET can be regarded as the equivalent of Fw of 7 nm Fh of
14 nm or 5 nm 20 nm rectangle FinFET, where k3 represents tri-gate (which is
identical to FinFET) and k4 represents to GAA FET.
This chapter starts with the discussion of simulation of 3D FinFET, including the
simulations of nFinFET, pFinFET Id–Vg and Id–Vd. First, the Predictive Technology
Model for FinFETs is explained based on the paper published by ARM Company in
ACM in 2012 [4], and the simulation process flow of its 3D FinFET is as shown in
Fig. 3.11.
102 3 3D FinFET with Lg = 15 nm and Lg = 10 nm Simulation
Gate
Silicon
MIGFET Cylindrical FET
π-gate FET
2D MOSFET
Multi -bridge/stacked
FinFET Ω-gate FET Nanowire FET
X
Z
Y
EY
EZ EX Drain
tsi EX
EZ
EY
wsi L
8
λ1
7 λ2
λ3
0
0 2 4 6 8 10 12 14 16 18 20 22
Tsi or Fh=Fw (nm)
NO
Ion, Ioff target achieved
END
104 3 3D FinFET with Lg = 15 nm and Lg = 10 nm Simulation
The first thing is to start building the structure of 3D FinFET by right clicking
commands on SDE tool to establish codes. The device length uses unit of nm.
Before the establishment of 3D device structure, the X–Y–Z coordinates and zero
point must be set firstly. Then, the “eight cuboids” form the basic structure of
FinFET and describe in the following order:
(1) Source contact (SC)
(2) Source (S)
(3) Gate oxide (Gox)
(4) Channel (channel)
(5) Drain (D)
(6) Drain contact (DC)
(7) Si Body (Body)
(8) STI buried oxide (Box) (Fig. 3.12).
DC X
D
Z 15
G
15
Y S Z
Lg
SC
15
15
(0.0.0)
Z
Si
Fh Channel
Fw
Y
Fig. 3.12 Eight cuboids’ structural diagram and 2D cross-sectional diagram of 3D FinFET
structure from Synopsys Sentaurus screen capture. The unit of length is nm
3.5 FinFET 3D Simulation 105
After the FinFET cuboids are all established for simulation, the control gate (G) can
be regarded as a conductor with voltage applied to the gate (Vg), and it is
equipotential. Then, the metal gate work function (WK) must be defined in de-
vise_dvs.cmd file, and @WK@ must be added as a variable of SDEVICE tool of
Sentaurus Workbench (SWB). The advantage of this approach is that multiple WK
variables can be assigned for adjustment of Vth of 3D FinFET as shown in
Fig. 3.15.
The physical and electrical properties can be examined during the analysis of
FinFET device by Sentaurus. The Sentaurus Visual tool is as shown in Fig. 3.16,
where the materials and their mesh can be selected to investigation via left toolbar.
In addition, Sentaurus Visual tool allows the selection of physical property to be
examined, such as energy band, electrons, and holes distributions. in toolbar. We
take the energy band diagram for illustration. First, using the Y-axis cutting
(Fig. 3.16), the 3D diagram converts to 2D diagram along Y-axis, and then the
X–Z cross section as shown in Fig. 3.17.
106 3 3D FinFET with Lg = 15 nm and Lg = 10 nm Simulation
(a)
DrainC
(Si)
Channel1 Drain x5
(Si) (Si)
x4
Source x3
(Si)
x2
SourceC
(Si) x1
(000)
(b)
z2
z1
y1y2
Fig. 3.13 Structure coordinated of a 3D FinFET structure and b Y–Z cross section of center of
FinFET
And then, the variation of physical properties along X-axis should be inspected
by cutting along Z-axis. The frame on the right side of the interface can be used to
selected the 1D doping concentration in Fig. 3.18, and 1D energy band diagram in
Fig. 3.19.
3.5 FinFET 3D Simulation 107
Fig. 3.14 Illustration of establishment of 3D cuboids (3D cuboid is formed by the diagonal from
A to B)
Fig. 3.15 Determination of variable values of @WK@ (4.65) metal work function, and gate
voltage and drain voltage @Vg@ (−1), and @Vd@ (−0.05 and −1) within SWB
Y cut
Fig. 3.16 3D structural diagram of n-type FinFET by Sentaurus Visual interface (readers can
select the required FinFET physical property diagram from the toolbar on the left)
108 3 3D FinFET with Lg = 15 nm and Lg = 10 nm Simulation
Z cut
Fig. 3.17 2D structural diagram of n-type FinFET by Sentaurus Visual interface via toolbar
Fig. 3.18 1D structural diagram of doping concentration of n-type FinFET by Sentaurus Visual
Fig. 3.19 Energy band diagram (Ec, Ev to X) of n-type FinFET by Sentaurus Visual interface
Fig. 3.20 Required simulation tools are shown in the workbench of 3D nFinFET TCAD
simulation
Lg=15nm D D
Fw =5nm
Fh =5nm G
STI SiO2
(3) Contact
(4) Doping
(5) Mesh
(6) Save
1. SDE -- devise_dvs.cmd
(define nm 1e-3)
(define Fw 5)
(define Fh 5)
(define Lg 15)
(define x1 LSDC)
(define x2 (+ x1 LSD))
(define x3 (+ x2 Lg))
(define x4 (+ x3 LSD))
(define x5 (+ x4 LSDC))
(define y1 Fw)
(define y2 (+ y1 Tox))
(define y3 (+ y2 10))
(define z1 Fh)
(define z2 (+ z1 Tox))
"ABA"
"ABA"
(sdegeo:set-current-contact-set "S")
(sdegeo:set-current-contact-set "D")
112 3 3D FinFET with Lg = 15 nm and Lg = 10 nm Simulation
(sdegeo:set-current-contact-set "G")
(sdegeo:set-current-contact-set "G")
(sdegeo:set-current-contact-set "G")
(sdegeo:set-current-contact-set "B")
(sdedr:define-refinement-size "Cha_Mesh" 5 5 5 1 1 1)
(sdedr:define-multibox-size "multiboxSizeChannel" 2 2 2 2 2 2)
; ----------- (6) Save (BND and CMD and rescale to nm) -----------;
(sde:assign-material-and-region-names (get-body-list) )
(define sde:scale-tdr-bnd
(sde:clear)
(sdegeo:set-default-boolean "XX")
(sdeio:read-tdr-bnd tdrin)
))
The “ABA” is important command to define gate insulator and channel region
for FinFET or other complex device structures. It defines the latter (or new) cuboid
replacing former (or old) cuboide in their overlapping region. Figure 3.22 illustrates
the “ABA” command results. On the other hand, “BAB” command can use for
former (or old) cuboide replacing latter (or new) cuboide in their overlapping
region.
(a)
Origin Box (Gox)
(b)
Origin Box (Gox)
Origin Origin
New New
box
box
box
Box
(Si)
Fig. 3.22 Illustration of “ABA” command. It is very useful in 3D FET, a is suitable for
gate-all-around FET, and b FinFET
3.5 FinFET 3D Simulation 115
2. SDEVICE -- dessis_des.cmd
Grid="@tdr@"
Plot="@tdrdat@"
Current="@plot@"
Output="@log@"
Electrode {
{ name="S" Voltage=0.0 }
{ name="D" Voltage=0.0 }
{ name="B" Voltage=0.0 }
Physics{
EffectiveIntrinsicDensity( OldSlotboom )
Recombination( SRH(DopingDep) )
Math{
Extrapolate
Derivatives
* Avalderivatives
RelErrControl
Digits=5
116 3 3D FinFET with Lg = 15 nm and Lg = 10 nm Simulation
ErRef(electron)=1.e10
ErRef(hole)=1.e10
Notdamped=50
Iterations=20
*Newdiscretization
Directcurrent
Method=ParDiSo
Parallel= 2
*-VoronoiFaceBoxMethod
NaturalBoxMethod
Plot{
eDensity hDensity
eCurrent hCurrent
eMobility hMobility
eVelocity hVelocity
eEnormal hEnormal
eQuasiFermi hQuasiFermi
SRH Auger
AvalancheGeneration
DonorConcentration AcceptorConcentration
Doping
eGradQuasiFermi/Vector hGradQuasiFermi/Vector
3.5 FinFET 3D Simulation 117
eEparallel hEparalllel
BandGap
BandGapNarrowing
Affinity
ConductionBand ValenceBand
Solve{
NewCurrentFile=""
Coupled(Iterations=100){ Poisson }
Quasistationary(
InitialStep=0.01 Increment=1.35
MinStep=1e-5 MaxStep=0.2
Quasistationary(
InitialStep=1e-3 Increment=1.35
MinStep=1e-5 MaxStep=0.05
3. INSPECT -- inspect_inc.cmd
#------------------------------------------------------------------------#
# * The transconductance : gm #
#------------------------------------------------------------------------ #
# idvgs=y(x) ; vgsvgs=x(x) ; #
proj_load "${out_file}.plt"
# ---------------------------------------------------------------------- #
# ---------------------------------------------------------------------- #
#....................................................................... #
#....................................................................... #
#................................................................ #
#................................................................ #
# ..................................................................... #
cv_display idvgs
# ---------------------------------------------------------------------- #
120 3 3D FinFET with Lg = 15 nm and Lg = 10 nm Simulation
# II) gm = maxslope((ID[VGS]) #
# ---------------------------------------------------------------------- #
ft_scalar VT $VT1
close $log_file
3D -nFinFET
Vd =1 V
Vd =0.05 V
S.S. = 67 mV/dec. @ Lg = 15 nm
Fig. 3.23 Id–Vg curve of 3D nFinFET simulation, some of the descriptions are added by
PowerPoint after snapshot by inspect tool
3D –nFinFET Mesh
Lg =15nm
Vd =1V
Vg =1V
3D – Electron concentration
Lg =15nm
Vd =1V
Vg =1V
2D – Electron concentration
Lg =15nm
Vd =1V
Vg =1V
3D – Electric Field
Lg =15nm
Vd =1V
Vg =1V
2D – Electric Field
Lg =15nm
Vd =1V
Vg =1V
3D – Electrostatic Potential
Lg =15nm
Vd =1V
Vg =1V
2D – Electrostatic Potential
Lg =15nm
Vd =1V
Vg =1V
Ec Lg=15nm Lg=15nm
Vg=1V Vg=1V
Vd =1V Vd =1V
Ec Channel
E (eV)
E (eV)
Ev Ev
S
Channel
Fig. 3.33 Required simulation tools are shown in the workbench of nFinFET simulation
3.5 FinFET 3D Simulation 127
1. SDE -- devise_dvs.cmd
(define nm 1e-3)
(define Fw 5)
(define Fh 5)
(define Lg 15)
(define x1 LSDC)
(define x2 (+ x1 LSD))
(define x3 (+ x2 Lg))
(define x4 (+ x3 LSD))
(define x5 (+ x4 LSDC))
(define y1 Fw)
(define y2 (+ y1 Tox))
(define y3 (+ y2 10))
(define z1 Fh)
(define z2 (+ z1 Tox))
"ABA"
"ABA"
(sdegeo:set-current-contact-set "S")
(sdegeo:set-current-contact-set "D")
(sdegeo:set-current-contact-set "G")
(sdegeo:set-current-contact-set "G")
(sdegeo:set-current-contact-set "G")
(sdegeo:set-current-contact-set "B")
; ----- Doping-----;
(sdedr:define-refinement-size "Cha_Mesh" 5 5 5 1 1 1)
(sdedr:define-multibox-size "multiboxSizeChannel" 2 2 2 2 2 2)
(sde:assign-material-and-region-names (get-body-list) )
(define sde:scale-tdr-bnd
(sdegeo:set-default-boolean "XX")
(sdeio:read-tdr-bnd tdrin)
2. SDEVICE -- dessis_des.cmd
File{
Grid="@tdr@"
Plot="@tdrdat@"
Current="@plot@"
Output="@log@"
}
132 3 3D FinFET with Lg = 15 nm and Lg = 10 nm Simulation
Electrode {
{ name="S" Voltage=0.0 }
{ name="D" Voltage=0.0 }
{ name="B" Voltage=0.0 }
Physics{
EffectiveIntrinsicDensity( OldSlotboom )
Recombination( SRH(DopingDep) )
Math{
Extrapolate
Derivatives
* Avalderivatives
RelErrControl
Digits=5
ErRef(electron)=1.e10
ErRef(hole)=1.e10
Notdamped=50
Iterations=20
3.5 FinFET 3D Simulation 133
*Newdiscretization
Directcurrent
Method=ParDiSo
Parallel= 2
*-VoronoiFaceBoxMethod
NaturalBoxMethod
Plot{
eDensity hDensity
eCurrent hCurrent
eMobility hMobility
eVelocity hVelocity
eEnormal hEnormal
eQuasiFermi hQuasiFermi
SRH Auger
AvalancheGeneration
DonorConcentration AcceptorConcentration
Doping
134 3 3D FinFET with Lg = 15 nm and Lg = 10 nm Simulation
eGradQuasiFermi/Vector hGradQuasiFermi/Vector
eEparallel hEparalllel
BandGap
BandGapNarrowing
Affinity
ConductionBand ValenceBand
Solve{
NewCurrentFile=""
Coupled(Iterations=100){ Poisson }
Quasistationary(
InitialStep=0.01 Increment=1.35
MinStep=1e-5 MaxStep=0.2
Quasistationary(
InitialStep=1e-3 Increment=1.35
MinStep=1e-5 MaxStep=0.05
3. INSPECT -- inspect_inc.cmd
#------------------------------------------------------------------------#
# * The transconductance : gm #
#------------------------------------------------------------------------#
# idvgs=y(x) ; vgsvgs=x(x) ; #
136 3 3D FinFET with Lg = 15 nm and Lg = 10 nm Simulation
proj_load "${out_file}.plt"
# ---------------------------------------------------------------------- #
# ---------------------------------------------------------------------- #
#....................................................................... #
#....................................................................... #
#................................................................ #
#................................................................ #
#...................................................................... #
# ..................................................................... #
cv_display idvgs
# ---------------------------------------------------------------------- #
# II) gm = maxslope((ID[VGS]) #
# ---------------------------------------------------------------------- #
ft_scalar VT $VT1
close $log_file
3D - nFinFET
Lg = 15 nm Vg = 1 V
Vg = 0.6 V
Vg = 0.2 V
Fig. 3.35 Required simulation tools are shown in the workbench of pFinFET simulation
140 3 3D FinFET with Lg = 15 nm and Lg = 10 nm Simulation
1. SDE -- devise_dvs.cmd
(define nm 1e-3)
(define Fw 5)
(define Fh 5)
(define Lg 15)
(define x1 LSDC)
(define x2 (+ x1 LSD))
(define x3 (+ x2 Lg))
(define x4 (+ x3 LSD))
(define x5 (+ x4 LSDC))
(define y1 Fw)
(define y2 (+ y1 Tox))
(define y3 (+ y2 10))
(define z1 Fh)
(define z2 (+ z1 Tox))
"ABA"
"ABA"
(sdegeo:set-current-contact-set "S")
(sdegeo:set-current-contact-set "D")
142 3 3D FinFET with Lg = 15 nm and Lg = 10 nm Simulation
(sdegeo:set-current-contact-set "G")
(sdegeo:set-current-contact-set "G")
(sdegeo:set-current-contact-set "G")
(sdegeo:set-current-contact-set "B")
; ----- Doping-----;
(sdedr:define-refinement-size "Cha_Mesh" 6 6 6 3 3 3)
(sdedr:define-multibox-size "multiboxSizeChannel" 2 2 2 1 1 1)
144 3 3D FinFET with Lg = 15 nm and Lg = 10 nm Simulation
(sde:assign-material-and-region-names (get-body-list) )
(define sde:scale-tdr-bnd
(lambda (tdrin sf tdrout)
(sde:clear)
(sdegeo:set-default-boolean "XX")
(sdeio:read-tdr-bnd tdrin)
2. SDEVICE -- dessis_des.cmd
File{
Grid="@tdr@"
Plot="@tdrdat@"
Current="@plot@"
Output="@log@"
Electrode {
{ name="S" Voltage=0.0 }
{ name="D" Voltage=0.0 }
{ name="B" Voltage=0.0 }
Physics{
EffectiveIntrinsicDensity( OldSlotboom )
Recombination( SRH(DopingDep) )
Math{
Extrapolate
Derivatives
* Avalderivatives
RelErrControl
Digits=5
ErRef(electron)=1.e10
146 3 3D FinFET with Lg = 15 nm and Lg = 10 nm Simulation
ErRef(hole)=1.e10
Notdamped=50
Iterations=20
*Newdiscretization
Directcurrent
Method=ParDiSo
Parallel= 2
*-VoronoiFaceBoxMethod
NaturalBoxMethod
Plot{
eDensity hDensity
eCurrent hCurrent
eMobility hMobility
eVelocity hVelocity
eEnormal hEnormal
ElectricField/Vector Potential SpaceCharge
eQuasiFermi hQuasiFermi
SRH Auger
AvalancheGeneration
DonorConcentration AcceptorConcentration
Doping
eGradQuasiFermi/Vector hGradQuasiFermi/Vector
eEparallel hEparalllel
3.5 FinFET 3D Simulation 147
BandGap
BandGapNarrowing
Affinity
ConductionBand ValenceBand
Solve{
NewCurrentFile=""
Coupled(Iterations=100){ Poisson }
Quasistationary(
InitialStep=0.01 Increment=1.35
MinStep=1e-5 MaxStep=0.2
Quasistationary(
InitialStep=1e-3 Increment=1.35
MinStep=1e-5 MaxStep=0.05
3. INSPECT -- inspect_inc.cmd
#------------------------------------------------------------------------#
# * The transconductance : gm #
#------------------------------------------------------------------------#
# idvgs=y(x) ; vgsvgs=x(x) ; #
# ---------------------------------------------------------------------- #
# ---------------------------------------------------------------------- #
#....................................................................... #
#....................................................................... #
#................................................................ #
#...................................................................... #
# ..................................................................... #
cv_display idvgs
# ---------------------------------------------------------------------- #
# II) gm = maxslope((ID[VGS]) #
# ---------------------------------------------------------------------- #
ft_scalar VT $VT1
close $log_file
The Id–Vg curve of the simulation result is as shown in Fig. 3.36, in which the
important parameters are SS at around 64 mV/dec., Vth at around −0.3 V, Isat at
1.29 10−5 A, and Ioff at 2.93 10−12 A as shown in Fig. 3.37. The structural
channel mesh, the electron concentration distributions of 3D and 2D structures,
electric field distributions, electric potential distributions, and the energy band
diagrams along the channel direction are as shown in Fig. 3.38, 3.39, 3.40, 3.41,
3.42, 3.43, 3.44 and 3.45 with the conditions of Lg = 15 nm, Vd = −1 V, and
Vg = −1 V, respectively.
Example 3.4 Comparison of different Fh (Fin height) with Lg = 10 nm
nFinFET
The following three main program code files are based on Synopsys
Sentaurus TCAD 2014 version.
We use gate length Lg = 10 nm nFinFET with Fw = 5 nm and Fh = 5, 10,
15, 20, 25, 30 and 35 nm at Vdd = 0.7 V are fir simulation (Figs. 3.46, 3.47 and
3.48).
3D - pFinFET
Vd = - 1 V
Vd = - 0.05 V
S.S. = 64 mV/dec. @ Lg = 15 nm
3D – pFinFET Mesh
Lg =15nm
Vd = -1V
Vg = -1V
3D – Hole concentration
Lg =15nm
Vd = -1V
Vg = -1V
2D – Hole concentration
Lg =15nm
Vd = -1V
Vg = -1V
3D – Electric Field
Lg =15nm
Vd = -1V
Vg = -1V
2D – Electric Field
Lg =15nm
Vd = -1V
Vg = -1V
3D – Electrostatic Potential
Lg =15nm
Vd = -1V
Vg = -1V
2D – Electrostatic Potential
Lg =15nm
Vd = -1V
Vg = -1V
Band Diagram
Channel
Lg =15nm
Vd =-1V
Vg =-1V
Fig. 3.46 Required simulation tools are shown in the workbench of Lg = 10 nm nFinFET with
different Fh simulations and Vdd = 0.7 V
Fig. 3.47 Electron current density and mesh plots of Lg = 10 nm nFinFET with Fw = 5 nm
and Fh = 35 m at Vdd = 0.7 V
3.5 FinFET 3D Simulation 157
Fig. 3.48 Electron mobility and mesh plots of Lg = 10 nm nFinFET with Fw = 5 nm and
Fh = 35 nm at Vdd = 0.7 V. The unstrained Si channel maximum mobility is around 130 cm2/Vs
158 3 3D FinFET with Lg = 15 nm and Lg = 10 nm Simulation
1. SDE -- devise_dvs.cmd
(define nm 1e-3)
(define Fw @Fw@)
(define Fh @Fh@)
(define Lg 10)
(define x1 LSDC)
(define x2 (+ x1 LSD))
(define x3 (+ x2 Lg))
(define x4 (+ x3 LSD))
(define x5 (+ x4 LSDC))
(define y1 Fw)
(define y2 (+ y1 Tox))
(define y3 (+ y2 10))
(define z1 Fh)
(define z2 (+ z1 Tox))
"ABA"
3.5 FinFET 3D Simulation 159
(sdegeo:set-current-contact-set "S")
(sdegeo:set-current-contact-set "D")
(sdegeo:set-current-contact-set "G")
(sdegeo:set-current-contact-set "G")
(sdegeo:set-current-contact-set "G")
(sdegeo:set-current-contact-set "B")
; -----Doping -----;
(sdedr:define-refinement-size "Cha_Mesh" 5 5 5 1 1 1)
(sdedr:define-multibox-size "multiboxSizeChannel" 2 2 2 2 2 2)
(sde:assign-material-and-region-names (get-body-list) )
(define sde:scale-tdr-bnd
(sde:clear)
(sdegeo:set-default-boolean "XX")
162 3 3D FinFET with Lg = 15 nm and Lg = 10 nm Simulation
(sdeio:read-tdr-bnd tdrin)
))
2. SDEVICE -- dessis_des.cmd
File{
Grid="@tdr@"
Plot="@tdrdat@"
Current="@plot@"
Output="@log@"
Electrode {
{ name="S" Voltage=0.0 }
{ name="D" Voltage=0.0 }
{ name="B" Voltage=0.0 }
}
3.5 FinFET 3D Simulation 163
Physics{
EffectiveIntrinsicDensity( OldSlotboom )
Recombination( SRH(DopingDep) )
Math{
Extrapolate
Derivatives
* Avalderivatives
RelErrControl
Digits=5
ErRef(electron)=1.e10
ErRef(hole)=1.e10
Notdamped=50
Iterations=20
*Newdiscretization
Directcurrent
Method=ParDiSo
Parallel= 2
*-VoronoiFaceBoxMethod
NaturalBoxMethod
}
164 3 3D FinFET with Lg = 15 nm and Lg = 10 nm Simulation
Plot{
eDensity hDensity
eCurrent hCurrent
eMobility hMobility
eVelocity hVelocity
eEnormal hEnormal
eQuasiFermi hQuasiFermi
SRH Auger
AvalancheGeneration
DonorConcentration AcceptorConcentration
Doping
eGradQuasiFermi/Vector hGradQuasiFermi/Vector
eEparallel hEparalllel
BandGap
BandGapNarrowing
Affinity
ConductionBand ValenceBand
}
3.5 FinFET 3D Simulation 165
Solve{
NewCurrentFile=""
Coupled(Iterations=100){ Poisson }
Quasistationary(
InitialStep=0.01 Increment=1.35
MinStep=1e-5 MaxStep=0.2
Quasistationary(
InitialStep=1e-3 Increment=1.35
MinStep=1e-5 MaxStep=0.05
3. INSPECT -- inspect_inc.cmd
#------------------------------------------------------------------------#
# * The transconductance : gm #
#------------------------------------------------------------------------#
# #
# idvgs=y(x) ; vgsvgs=x(x) ; #
# #
proj_load "${out_file}.plt"
3.5 FinFET 3D Simulation 167
# ---------------------------------------------------------------------- #
# ---------------------------------------------------------------------- #
#....................................................................... #
#....................................................................... #
#................................................................ #
#................................................................ #
#...................................................................... #
# ..................................................................... #
cv_display idvgs
# ---------------------------------------------------------------------- #
# II) gm = maxslope((ID[VGS]) #
# ---------------------------------------------------------------------- #
ft_scalar VT $VT1
ft_scalar gmax $gm
close $log_file
The important key performance index (KPI), Vth, Ioff, Ion and SS, Isat, with
respect to different Fh is as shown in Fig. 3.49 . The Id–Vg of simulation result is as
shown in Fig. 3.50. It is shown in the figure that smaller Fh will lead to smaller Ioff
and SS, indicating better gate control capability. Again, it has been mentioned
previously that the smaller Natural length will be preferred. In general, Lg > 5–10
times of k. Thus, smaller Fh will lead to smaller value of natural length.
Example 3.5 Si1−xGex pFinFET with Lg = 15 nm and HfO2 as gate insulator
Material selection for sub-7-nm technology node:
(a) Silicon
Silicon is the most important semiconductor material of the world. It is fairly
abundant in nature, easy to be process, and equipped with characteristics of nFET
and pFET with excellent stability and very low cost. These characteristics have
made silicon the favorite material in semiconductor industry until 10-nm nodes or
beyond.
Lg=10nm, nFinFET
Fh =35nm
Vd =0.7V
Step=5nm
Fh =5nm
Fig. 3.51 SWB shows Si1−xGex pFinFET with Lg = 15 nm, and set Mole-Fraction x as xF from
0 to 1.0 as step 0.1, where silicongermanium (x = 0) = silicon and silicongermanium
(x = 1) = germanium
Lg=15nm D
Ge
HfO2
Ge
(b) Germanium
Germanium is also equipped with many advantages. For example, it is equipped
with higher electron mobility and hole mobility than silicon of Table 3.4, and it can
be used for high-frequency field. These advantages plus the similar fabrication
process to silicon have made Germanium a possible candidate for applications at
sub-10-nm node. The Si1−xGex material has higher mobility and is compatible to
current Si-based FinFET process. Therefore, for Example 3.5, we study the Si1
−xGex FinFET by using high-k material HfO2 of 3 nm. The Si1−xGex Ge molecular
fraction X changes from 0 to 1, step 0.1, which is shown in Fig. 3.51.
In Sentaurus Device Mole-Fraction Materials section (Sentaurus™ Device
User Guide J-2014.09):
3.5 FinFET 3D Simulation 171
Fig. 3.53 Results electric properties of Si1−xGex pFinFET with Lg = 15 nm, and set
Mole-Fraction x as xF from 0 to 1.0 as step 0.1, where silicongermanium (x = 0) = silicon and
silicongermanium (x = 1) = germanium. The important parameters Vth, Ioff, Isat, and SS are shown
Ec
E (eV)
Ev
Y position (um)
172 3 3D FinFET with Lg = 15 nm and Lg = 10 nm Simulation
Vd =-1.0V
X=0.5
0.4
0.3
Is (A)
Lg=15nm
Si 1-xGe x
pFinFET 0.2
X=0.1
Vg (V)
Fig. 3.55 Extraction Is–Vg curves from Fig. 3.53 Si1−xGex pFinFET with Lg = 15 nm and
different x fractions from 0.1 to 0.5. The higher x value has higher Ion and Ioff
0.60
In0.53Ga0.47As InP
0.58
Ge GaAs
0.56
Si
0.54
0 0.5 1.0 1.5 2.0
Band gap E g (eV)
3.5 FinFET 3D Simulation 173
Lg=100nm
InGaAs
Fh =30nm
Fw =50nm HfO2/Al2O3 D
InGaAs
InP
STI SiO2
Si substrate
Fig. 3.57 In1−xGaxAs nFinFET structure with stacked gate insulator HfO2/Al2O3 (top/bottom) of
3/3 nm
Fig. 3.58 SWB shows In1−xGaxAs nFinFET with Lg = 100 nm, and set Mole-Fraction x as xF
from 0 to 1.0 as step 0.2, and 0.47. For x = 0.47, In1−xGaxAs is In0.53Ga0.47As
174 3 3D FinFET with Lg = 15 nm and Lg = 10 nm Simulation
3. If these previous strategies are unsuccessful, Sentaurus Device uses the built-in
defaults that follow.
# Ge(x)Si(1-x)
# Al(x)Ga(1-x)As
# In(1-x)Al(x)As
# In(1-x)Ga(x)As
# Ga(x)In(1-x)P
# InAs(x)P(1-x)
# GaAs(x)P(1-x)
# Hg(1-x)Cd(x)Te
3.5 FinFET 3D Simulation 175
# In(1-x)Ga(x)As(y)P(1-y)
1. SDE tool
;---------- example 3.5 SiGex pFinFET with Lg=15nm and HFO2 ---------;
(define nm 1e-3)
(define Fw 5)
(define Fh 5)
(define Lg 15)
(define x1 LSDC)
(define x2 (+ x1 LSD))
(define x3 (+ x2 Lg))
(define x4 (+ x3 LSD))
(define x5 (+ x4 LSDC))
(define y1 Fw)
176 3 3D FinFET with Lg = 15 nm and Lg = 10 nm Simulation
(define y2 (+ y1 THfO2))
(define y3 (+ y2 10))
(define z1 Fh)
(define z2 (+ z1 THfO2))
"ABA"
"ABA"
(sdegeo:set-current-contact-set "S")
(sdegeo:set-current-contact-set "D")
(sdegeo:set-current-contact-set "G")
(sdegeo:set-current-contact-set "G")
(sdegeo:set-current-contact-set "G")
(sdegeo:set-current-contact-set "B")
(sdedr:define-refinement-size "Cha_Mesh" 5 5 5 1 1 1)
(sdedr:define-multibox-size "multiboxSizeChannel" 2 2 2 2 2 2)
(sde:assign-material-and-region-names (get-body-list) )
(define sde:scale-tdr-bnd
(sde:clear)
(sdegeo:set-default-boolean "XX")
(sdeio:read-tdr-bnd tdrin)
)
(sde:scale-tdr-bnd "n@node@_nm.tdr" nm "n@node@_bnd.tdr")
Lg=100nm D
In0.53Ga0.47As
Si substrate
First, in SEVICE tool must include parameter files, SiliconGermanium and HfO2. The
dessis_des.cmd is identical to example 3.1, only add following text
Physics(material="SiliconGermanium"){
MoleFraction(xFraction=@xF@)
# In(1-x)Ga(x)As
Ev
Drain
Vd=1V
Ev
Fig. 3.60 1D energy band diagram of In0.53Ga0.47As nFinFET for a X and b Y direction
182 3 3D FinFET with Lg = 15 nm and Lg = 10 nm Simulation
(a) (b)
Ec
Vd= 1V
In0.53Ga0.47As Lg=100nm
Ev Vd=0.05V
Eg=0.7eV In 0.53 Ga 0.47 As
Si nFinFET
Substrate InP
Eg=1.12eV
E(eV)
Eg=1.3eV
Id (A)
SS ~ 65 mc/dec.
Vg=1V, Vd=1V
Fig. 3.61 a 1D energy band diagram of In0.53Ga0.47As nFinFET for Z direction. b Id–Vg of
In0.53Ga0.47As nFinFET
References 183
References
1. C.C. Hu, in Modern Semiconductor Devices for Integrated Circuits (PEARSON, 2010)
2. C.H. Jan, U. Bhattacharya, R. Brain, S.-J. Choi, G. Curello, G. Gupta, W. Hafez, M. Jang,
M. Kang, K. Komeyli, T. Leo, N. Nidhi, L. Pan, J. Park, K. Phoa, A. Rahman, C. Staus,
H. Tashiro, C. Tsai, P. Vandervoorn, L. Yang, J.-Y. Yeh, P. Bai, A 22 nm SoC platform
technology featuring 3-D tri-gate and high-k/metal gate, optimized for ultra low power, high
performance and high density SoC applications. Tech. Digest IEDM 3.1.1 (2012)
3. S. Sinha, B. Cline, G. Yeric, V. Chandra, Y. Cao, Design benchmarking to 7 nm with FinFET
predictive technology models. ISLPED 15 (2012)
4. M. Bohr, Opening New Horizons: 14 nm Process Technology (Intel IDF, 2014)
5. J.P. Colinge, in FinFETs and Other Multi-Gate Transistors (Springer, 2007)
6. N. Waldron, C. Merckling, W. Guo, P. Ong, L. Teugels, S. Ansar, D. Tsvetanova, F. Sebaai,
D.H. van Dorp, A. Milenin, D. Lin, L. Nyns, J. Mitard, A. Pourghaderi, B. Douhard,
O. Richard, H. Bender, G. Boccardi, M. Caymax, M. Heyns, W. Vandervorst, K. Barla,
N. Collaert, A.V.Y. Thean, An InGaAs/InP quantum well finfet using the replacement fin
process integrated in an RMG flow on 300 mm Si substrates. VLSI Tech. Symp. 1 (2014)
Chapter 4
Inverter and SRAM of FinFET
with Lg = 15 nm Simulation
Consider the CMOS Inverter shown in Fig. 4.1a. The IV curve of nFET is as shown
on the right half of 4.1b. Assume that the pFET has identical (symmetric) IV as
plotted on the left half of the figure. From (a), the Vds of the pFET and nFET are
related to Vout by VdsN = Vout and VdsP = Vout − 2 V. Therefore, the two halves of
(b) can be replotted in (c) using Vout as the common variable. For example, at
Vout = 2 V in (c), VdsN = 2 V and VdsP = 0 V.
The two Vin = 0 V curves in Fig. 4.1c intersect at Vout = 2 V. This means
Vout = 2 V when Vin = 0. This point is recorded in Fig. 4.2. The two Vin = 0.5 V
curves intersect at around Vout = 1.9 V. The two Vin = 1 V curves intersect at
Vout = 1 V. All the Vin/Vout pairs are represented by the curve in Fig. 4.2, which is
the voltage transfer characteristic of Inverter or voltage transfer curve (VTC).
The VTC provides digital circuit with important noise margin. Vin can be anywhere
from 0 V to Vth of nFET while resulting in the ideal Vout = Vdd. Similarly, Vin can
be anywhere between 2 and 2 V plus Vth of pFET while resulting in the ideal
Vout = 0 V.
Idd (mA)
(a) 2V (c) 0V 2V
0.2
pFET 1.5 V
0.5 V
0.1
V in Vout
1V 1V
nFET
Vout (V)
0 0.5 1.0 1.5 2.0
0V
(b)
Idd (mA ) Vin = 2V
Vin = 0V
0.2
pFET nFET
Vin = 0.5V Vin = 1.5V
0.1
Vin = 1V Vin = 1V
Vin = 1.5V Vin = 0.5V
1.5
1.0
0.5
Vin
0 0.5 1.0 1.5 2.0
4.1 Voltage Transfer Curve of Inverter 187
Therefore, perfect “0” and “1” outputs can be produced by somewhat corrupted
inputs. This regenerative property allows complex logic circuits to function prop-
erly in the face of inductive and capacitive noises and IR drops in the signal lines.
A VTC with a narrow and steep middle region will maximize the noise tolerance.
Device characteristics which can be used to generate ideal VTC include greater gm,
low leakage current in the off-state, and small DIds/DVds in saturation region. The
two device characteristics will be further discussed in the next section.
During the operation of an ideal circuit, the transition region of VTC should be
located at or near Vin = Vdd/2. For achieving the symmetry, the IV curves of nFET
and pFET in Fig. 4.1b must be matched while being folded. This is achieved by
choosing the width (W) of transistors, W value of pFET greater than the W value of
nFET. Generally speaking, the ratio of Wp/WN is around 3 in order to compensate
the ratio of lps/lns = 3 in Lg = 15 nm FinFET.
The propagation delay is the delay time sd required by the signal to propagate from
one logic gate to the next identical logic gate as shown in Fig. 4.3.
sd is the average of falling delay (rising V1 pulling down the output, V2) and
rising delay (falling V2 pulling up the output, V3). The propagation delay of the
Inverter is expressed below:
(b)
Vdd V2 V3
2τd
V1
0 t
188 4 Inverter and SRAM of FinFET with Lg = 15 nm Simulation
CVdd 1 1
sd þ ð4:1Þ
4 IonN IonP
IonN is taken at Vgs = Vdd and IonP taken at Vgs = −Vdd. They are called the
on-state current of the nFET and the pFET.
Ion Idsat maxjvgs j ð4:2Þ
1
sD ¼ ðsF pull down delay þ sR pull up delayÞ ð4:3Þ
2
CVdd
pull down delay sF
2IonN
ð4:4Þ
CVdd
pull up delay sR
2IonP
The total delay refers to the time required for a conducting transistor to provide
an Ion current to change the output by Vdd/2 (not Vdd) in Fig. 4.3b. The charge
drained from C by the FET during the delay is CVdd/2. Therefore, the delay is
sd = Q/I = CVdd/2Ion. We can regard sd as RC delay, and Vdd/2Ion as transistor
switching resistance of the transistor. For maximizing the circuit operation speed,
Ion must be maximized, and the electric properties of pFET and nFET must be in
perfect symmetry.
As CMOS IC technology scales down to the 14-nm node, the FinFET transistor
technology has proven its superior capability to enable very aggressive and fol-
lowing Moore’s law. The 14-nm technology with a wide range of system-on-chip
(SoC) products, including tablets, smart phones, ASIC, embedded,
Internet-of-Things, baseband, and RF products [3].
The high performance (HP), standard performance (SP), and ultra-low power
(ULP) logic transistors are fully compatible with different SoC applications show
4.3 CMOS Id–Vg Matching Diagram for High-Performance Transistors 189
The nFinFET and pFinFET program code file is identical to Example 3.1 (or
Table 4.1, 4.1a) and Example 3.3 (or Table 4.1, 4.1c Fw/Fh = 16 nm/5 nm) of
Chap. 3, respectively. Example 4.1 is hereby combined nFinFET and pFinFET to
form the entire SDE codes of Inverter.
1e-4
Vd=-1V Vd=1V
1e-5
1e-7
1e-8
Id (A)
1e-13
-1.2 -0.9 -0.6 -0.3 0.0 0.3 0.6 0.9 1.2
Vg (V)
Fig. 4.4 Id–Vg characteristic in which the electric properties of nFinFET (Example 3.1) and
pFinFET (Example 3.3) must be perfectly matched for the inverter to be equipped with excellent
properties
190 4 Inverter and SRAM of FinFET with Lg = 15 nm Simulation
Table 4.1 Electric properties matching parameters of nFinFET and pFinFET extracted from
Fig. 4.4
(nm)
Example Vd = 1 V Fw/Fh WF (V) Vt (V) SS (mV/dec) Isat (A)
4.1a nFinFET 5/5 4.48 0.3020 65.8 2.685E−5
4.1b pFinFET 5/5 4.82 −0.3038 65.2 1.853E–5
4.1c pFinFET 16/5 4.80 −0.2939 68.3 2.69e−05
The width ratio is Wp/WN = 16 nm/5 nm–3 for ideal CMOSFET matching
Fig. 4.5 Required simulation tools are shown in the workbench of inverter simulation on SWB
The following three main tools and their code files are based on Synopsys
Sentaurus TCAD 2014 version (tool ! *.cmd)
SDE ! devise_dvs.cmd, SDEVICE ! dessis_des.cmd, and INSPECT !
inspect_inc.cmd.
Figure 4.5 is simulation tools of SWB for Inverter simulation.
Figure 4.7 is the simulated input and output electrical properties of Inverter
based on 3D FinFET with Lg = 15 nm.
1. SDE – devise_dvs.cmd
The SDE CODE of Examples 3.1 and 3.3 is substituted into nFET and
pFET, respectively. The first two nodes are for nFET, and the following two
nodes are for pFET.
Therefore, here we only need to focus on the following SDEVICE program
codes. Here, we use SDEVICE default library example.
2. SDVICE – dessis_des.cmd
4.4 [Example 4.1] Inverter of 3D FinFET with Lg = 15 nm 191
}}
File{
Output = "@log@"
}
Plot{
*----------------------------Density and Currents, etc
eDensityhDensity
TotalCurrent/Vector eCurrent/Vector hCurrent/Vector
eMobilityhMobility
eVelocityhVelocity
eQuasiFermihQuasiFermi
*----------------------------Fields and charges
ElectricField/Vector Potential SpaceCharge
*----------------------------Doping Profiles
Doping DonorConcentrationAcceptorConcentration
*----------------------------Generation/Recombination
SRH Auger
* AvalancheGenerationeAvalancheGenerationhAvalancheGeneration
*---------------------------Driving forces
eGradQuasiFermi/VectorhGradQuasiFermi/Vector
eEparallelhEparalllel
*---------------------------Band structure/Composition
BandGap
BandGapNarrowing
Affinity
ConductionBandValenceBand
}
Math (Region="Channel") {Nonlocal(-Transparent)
}
Math{
Extrapolate
Derivatives
* Avalderivatives
RelErrControl
4.4 [Example 4.1] Inverter of 3D FinFET with Lg = 15 nm 193
Digits=5
ErRef(electron)=1.e10
ErRef(hole)=1.e10
Notdamped=50
Iterations=20
*Newdiscretization
Directcurrent
Method=ParDiSo
Parallel= 2
*-VoronoiFaceBoxMethod
NaturalBoxMethod
}
System{
Vsource_pset VVDD (vdd 0) { dc = 0 }
Vsource_pset VGND (gnd 0) { dc=0 }
Vsource_pset VVIN (in 0) { pulse = ( 0 @Vdd@ 0.3e-10 0.02e-10
0.02e-100.3e-10 3000 )} (1) (2) (3) (4)
(5) (6) (7)
Note: The aforementioned Vin input square wave diagram is as shown on the left
of Fig. 4.6, and the Inverter circuit diagram is on the right.
(1). (2). … (7) are the program codes for definition of input square wave.
194 4 Inverter and SRAM of FinFET with Lg = 15 nm Simulation
Fig. 4.6 Vin input square wave diagram and inverter circuit diagram. 1 lowest voltage, 2 Vdd,
3 delay time, 4 rising time, 5 falling time, 6 duration time, and 7 period time
Solve{
Coupled(Iterations=150){ Poisson }
Coupled{ Poisson}
Coupled{ Poisson Electron Hole Contact Circuit }
Quasistationary(
InitialStep=1e-3 Increment=1.35
MinStep=1e-8 MaxStep=0.05
Goal{ Parameter=VVDD.dc Voltage= @Vdd@ } )
{Coupled{nmos1.poisson nmos1.electron nmos1.hole
nmos1. nmos1.contact pmos1.poisson pmos1.electron pmos1.hole
pmos1.pmos1.contact circuit }}
NewCurrentfile = "TR_"
Transient(
InitialTime=0
FinalTime=0.8e-10
InitialStep=1e-12
MaxStep=1e-11
Minstep=1.e-18
Increment=1.1)
{ Coupled{nmos1.poisson nmos1.electron nmos1.hole nmos1.
nmos1.contactpmos1.poissonpmos1.electron pmos1.hole pmos1. pmos1.contact
circuit }}}
* ---------------------------------------- END ------------------------------------------*
4.4 [Example 4.1] Inverter of 3D FinFET with Lg = 15 nm 195
Inverter
Lg=15nm
Vdd =1V
tHL=5.0E-13 s
tLH=8.1E13 s
Figure 4.7 is the simulated input and output electric properties of Inverter of 3D
FinFET with Lg = 15 nm.
stored charges can gradually disappear as time goes by, thus it will require certain
refreshing to keep the data stored in the capacitors.
Among all memories in Table 4.2, SRAM provides the fastest operating speed.
However, it will take six transistors to store one bit of data, thus leading to highest
cost per bit. When the processor speed is an important consideration, SRAM is
often used as the cache memory embedded in the processor.
In CMOS VLSI designs, the most commonly used SRAM storage element is
bistable latch consisting of two cross-coupled CMOS Inverters shown in Fig. 4.8. It
can be built using a standard CMOS logic fabrication process. Inverter 1 consists of
nFinFET Q1 and pFinFET Q3 while Inverter 2 consists of nFinFET Q2 and
pFinFET Q4. The two stable states can be readily recognized by plotting the transfer
curves of the two Inverters back to back, as illustrated in Fig. 4.9, often referred to
as the “butterfly curve” plot of a pair of cross-coupled Inverters.
In Fig. 4.8, one of the Inverters has its input at high and output at low, while the
other Inverter has its input at low and output at high. The first Inverter, with its
output at low, keeps the second Inverter in the state described above, and vice versa.
Thus, a CMOS SRAM storage element has two stable states: one at intersection A
of the two Inverter transfer curves in Fig. 4.9 with V1 = Vin2 = Vdd, and the other at
the intersection B with V2 = Vin1 = Vdd. The two stable states can be interpreted as
logical “0” and “1”. Here, we designate logical “1” as V1 = 0 and V2 = Vdd, i.e.,
point B, and logical “0” as V1 = Vdd and V2 = 0, i.e., point A. A bistable latch will
remain in one of its two stable states until it is forced by an external signal to flip to
the other stable state.
The most commonly used SRAM cell is a six-transistor cell consisting of two
cross-coupled CMOS Inverters and two access transistors. The circuit schematic for
CMOS SRAM cell is shown in Fig. 4.8. The cross-coupled Inverters are connected
to two bitlines, BLT (bitline true) and BLC (bitline complement), through
4.6 SRAM Operation 197
BLT BLC
(b)
n-channel access transistors Q5 and Q6. The access transistors are controlled by the
word line (WL) voltage. In the standby mode, WL is kept low (VWL = 0 V), thus
turning off the access transistors and isolating the bitlines from the cross-coupled
Inverters pair.
The two switch transistors, Q5 and Q6, are connecting the output of Inverter to
the bitline. For reading the saved data (by determining the state of Inverter), the WL
of the selected unit will be raised to high potential to turn on the access transistor.
A sensitive sensing amplifier circuit will compared the voltage differences of BLT
and BLC to determine the saved state. In order to write the low state “0” into the
unit on the left, BLT will be set at low potential and BLC will be set at high
potential.
The example of SRAM Read operation: SRAM operation for reading “0” is
as shown in Fig. 4.10. Where V1 = Vdd = 1 V, and V2 = 0 V.
The reading and writing of SRAM will be executed via conduction of word line
(WL) (“Logic 1”) (conduction between node Q5 and Q6). Here, we assume the data
saved in the V2 of SRAM unit is “Logic 0”.
The reading operation process is as shown below:
198 4 Inverter and SRAM of FinFET with Lg = 15 nm Simulation
A
Logical “0”
Inverter 1
Inverter 2
B
Logical “1”
Fig. 4.9 Butterfly plot for two cross-coupled CMOS Inverters. The transfer curve of inverter 1
(solid) is plotted as V1 versus Vin1, and inverter 2 (dashed) as Vin2 versus V2
Fig. 4.10 a Relationship between voltage and current during CMOS SRAM operation for reading
“0”, b the electric potential analysis of Node V1 and Node V2 during CMOS SRAM operation for
reading “0”
4.6 SRAM Operation 199
(1) Before reading: the node voltage V1 is logic “1” (Vdd) and V2 is logic “0”
(0 V).
(2) The parasitic capacitors of the two lines of BLT and BLC are pre-charged to
logic 1 such that the node voltage is VBLT = VBLC = Vdd, than selected WL
turn on.
(3) Q5 transistor is not conducting because the voltage on two terminals
VBLT = V1 = Vdd.
(4) In Q6 transistor, the capacitor CV2 occurs charge sharing because the voltage
on two terminals VBLC = Vdd with V2 = 0 V.
(5) As for the bitline (BL), VBLC will be less than VBLT due to charge sharing and
capacitive voltage division effects. When the difference between them is as
high as ΔV (around 0.1–0.2 V), it will trigger the sense amplifier to amplify
the signal difference. After it is converted into the output, it will be sent to the
data buffer to complete the operation of reading “0”. Even though there can be
interference during reading, the positive feedback of latch can restore V2 to
0 V, which is logic 0.
(6) After the operation is completed, WL will turn off to 0 V such that this SRAM
will be disconnected from BLs, thus ending the entire reading operation.
SRAM Write operation is as described below: it is the example of SRAM
operation for writing “1” (from the original state of “0” to “1”) is as shown in
Fig. 4.11.
The reading and writing of SRAM will be executed via conduction of word line
(WL) (“Logic 1”) (conduction between node Q5 and Q6). Here, we assume the data
saved in the V2 of SRAM unit is “Logic 0”, and now “Logic 1” must be written.
The writing operation process is as shown below:
(a)
(b)
dd
0 0
dd
Vdd dd
BLC
BLC
Vdd dd
Fig. 4.11 a Relationship between voltage and current of CMOS SRAM during the operation of
writing “1” b the electric potential analysis of Node V1 and Node V2 during CMOS SRAM
operation for writing “1”
200 4 Inverter and SRAM of FinFET with Lg = 15 nm Simulation
(1) Before writing: the node voltage V1 is logic “1” (Vdd) and V2 is logic “0”
(0 V).
(2) The parasitic capacitor of node BLT is reduced to logic “0”, and the parasite
capacitor of node BLC is pre-charged to logic “1”, such that node voltage
VBLT = 0 V; VBLC = Vdd, than selected WL turn on.
(3) As for transistors Q5 and Q3, there is current flowing through Q3 into Q5.
(4) As for transistors Q6 and Q2, capacitor CV2 can be rapidly charged via Q6.
(5) As for the bitlines (BLT, BLC), the voltage on BLT forces V1 to “0”, while the
voltage on BLC forces V2 to Vdd, thus writing a logic “1” to the cell.
(6) After the operation is completed, WL will turn off to 0 V such that this SRAM
will be disconnected from BLs, thus ending the entire writing operation.
The identical program codes of nFinFET of Example 3.1 (or Table 4.1, 4.1a) and
pFinFET of Example 3.3 (or Table 4.1, 4.1c Fw/Fh = 16 nm/5 nm) to form the
SRAM as shown below. SRAM is composed of 6 FinFETs (4 nFinFETs and 2
pFinFETs) as shown in Fig. 4.8a.
The following three main tools and their code files are based on Synopsys
Sentaurus TCAD 2014 version (tool ! *.cmd)
SDE ! NPNPNN-FET six devise_dvs.cmd, SDEVICE ! dessis_des.cmd,
and INSPECT ! inspect_inc.cmd.
1. SDE – devise_dvs.cmd
As shown in Fig. 4.12 workbench, in this example with a sequence of
(nFET-pFET: Inverter1), (nFET-pFET: Inverter2), and nFET-nFET (access tran-
sistors) SRAM, we only explain the following program codes of SDEVICE. Here,
we use SDEVICE default library example.
2. SDEVISE – dessis_des.cmd
Fig. 4.12 Required simulation tools are shown in the workbench for SRAM simulation
4.7 [Example 4.2] Simulation of SRAM of 3D FinFET with Lg = 15 nm 201
EffectiveIntrinsicDensity(oldSlotboom )
Recombination( SRH(DopingDep) )
}}
Device NMOS2 {
Electrode{
{ Name="S" Voltage=0.0 }
{ Name="D" Voltage=0.0 }
{ Name="G" Voltage=0.0 Workfunction= @WKN@}
*{ Name="B" Voltage=0.0 }
}
File{
Grid = "@tdr|-6@"
Plot = "@tdrdat@"
Current = "@plot@"
*Output= "@log@"
}
Physics{
Mobility(DopingDepHighFieldSaturationEnormal )
EffectiveIntrinsicDensity(OldSlotboom )
Recombination( SRH(DopingDep) )
}}
Device PMOS2{
Electrode{
{ Name="S" Voltage=0.0 }
{ Name="D" Voltage=0.0 }
{ Name="G" Voltage=0.0 Workfunction= @WKP@ }
*{ Name="B" Voltage=0.0 }
}
File{
Grid = "@tdr|-4@"
Plot = "@tdrdat@"
Current = "@plot@"
*Output = "@log@"
}
4.7 [Example 4.2] Simulation of SRAM of 3D FinFET with Lg = 15 nm 203
Physics{
Mobility(DopingDepHighFieldSaturationEnormal )
EffectiveIntrinsicDensity(oldSlotboom )
Recombination( SRH(DopingDep) )
}}
Device NMOS3 {
Electrode{
{ Name="S" Voltage=0.0 }
{ Name="D" Voltage=0.0 }
{ Name="G" Voltage=0.0 Workfunction= @WKN@}
* { Name="B" Voltage=0.0 }
}
File{
Grid = "@tdr|-2@"
Plot = "@tdrdat@"
Current = "@plot@"
*Output= "@log@"
}
Physics{
Mobility(DopingDepHighFieldSaturationEnormal )
EffectiveIntrinsicDensity(OldSlotboom )
Recombination( SRH(DopingDep) )
}}
Device NMOS4{
Electrode{
{ Name="S" Voltage=0.0 }
{ Name="D" Voltage=0.0 }
{ Name="G" Voltage=0.0 Workfunction= @WKN@ }
* { Name="B" Voltage=0.0 }
}
File{
Grid = "@tdr@"
Plot = "@tdrdat@"
Current = "@plot@"
204 4 Inverter and SRAM of FinFET with Lg = 15 nm Simulation
*Output = "@log@"
}
Physics{
Mobility(DopingDepHighFieldSaturationEnormal )
EffectiveIntrinsicDensity(oldSlotboom )
Recombination( SRH(DopingDep) )
}}
File{
Output = "@log@"
}
Plot{
*--Density and Currents, etc
eDensityhDensity
TotalCurrent/Vector eCurrent/Vector hCurrent/Vector
eMobilityhMobility
eVelocityhVelocity
eQuasiFermihQuasiFermi
*--Fields and charges
ElectricField/Vector Potential SpaceCharge
*--Doping Profiles
Doping DonorConcentrationAcceptorConcentration
*--Generation/Recombination
SRH Auger
* AvalancheGenerationeAvalancheGenerationhAvalancheGeneration
*--Driving forces
eGradQuasiFermi/VectorhGradQuasiFermi/Vector
eEparallelhEparalllel
*--Band structure/Composition
BandGap
BandGapNarrowing
Affinity
ConductionBandValenceBand }
Math{
4.7 [Example 4.2] Simulation of SRAM of 3D FinFET with Lg = 15 nm 205
Extrapolate
Derivatives
* Avalderivatives
RelErrControl
Digits=5
ErRef(electron)=1.e10
ErRef(hole)=1.e10
Notdamped=50
Iterations=20
*Newdiscretization
Directcurrent
Method=ParDiSo
Parallel= 2
*-VoronoiFaceBoxMethod
NaturalBoxMethod }
System{
Vsource_psetvdd (dd 0) { dc = 0.0 }
Vsource_psetvwl (T 0) { dc = 0.0 }
Vsource_psetvb (L 0) { dc = 0.0 }
Vsource_psetvbl (R 0) { dc = 0.0 }
Vsource_pset vin (VinL 0) { dc = 0.0 }
NMOS1 nmos1( "S"=0 "D"=VinR "G"=VinL )
NMOS2 nmos2( "S"=0 "D"=VinL "G"=VinR )
PMOS1 pmos1( "S"=dd "D"=VinR "G"=VinL )
PMOS2 pmos2( "S"=dd "D"=VinL "G"=VinR )
NMOS3 nmos3( "S"=VinR "D"=L "G"=T )
NMOS4 nmos4( "S"=R "D"=VinL "G"=T )
Plot "n@node@_sys_des.plt" (time() v(VinL) v(dd) v(T) v(L) v(R) v(VinR))}
Solve{
NewCurrentFile="init"
Coupled(Iterations=250){ Poisson }
Coupled{ Poisson Electron Hole Contact Circuit }
Quasistationary(
206 4 Inverter and SRAM of FinFET with Lg = 15 nm Simulation
InitialStep=1e-3 Increment=1.35
MinStep=1e-12 MaxStep=0.05
Goal{ Parameter=vdd.dc Voltage=@Vdd@ }
Goal{ Parameter=vwl.dc Voltage= @Vdd@ }
Goal{ Parameter=vb.dc Voltage= @Vdd@ }
Goal{ Parameter=vbl.dc Voltage= @Vdd@}
){ Coupled{ nmos1.poisson nmos1.electron nmos1. nmos1.contact
nmos2.poisson nmos2.electron nmos2. nmos2.contact
nmos3.poisson nmos3.electron nmos3. nmos3.contact
nmos4.poisson nmos4.electron nmos4. nmos4.contact
pmos1.poisson pmos1.hole pmos1. pmos1.contact
pmos2.poisson pmos2.hole pmos2. pmos2.contact
circuit }
}
Quasistationary(
InitialStep=1e-3 Increment=1.35
MinStep=1e-12 MaxStep=0.05
Goal{ Parameter=vdd.dc Voltage=@Vdd@ }
){ Coupled{ nmos1.poisson nmos1.electron nmos1. nmos1.contact
nmos2.poisson nmos2.electron nmos2. nmos2.contact
nmos3.poisson nmos3.electron nmos3. nmos3.contact
nmos4.poisson nmos4.electron nmos4. nmos4.contact
pmos1.poisson pmos1.hole pmos1. pmos1.contact
pmos2.poisson pmos2.hole pmos2. pmos2.contact
circuit }
}
Quasistationary(
InitialStep=1e-3 Increment=1.35
MinStep=1e-12 MaxStep=0.05
Goal{ Parameter=vwl.dc Voltage= @Vdd@ }
){ Coupled{ nmos1.poisson nmos1.electron nmos1. nmos1.contact
nmos2.poisson nmos2.electron nmos2. nmos2.contact
nmos3.poisson nmos3.electron nmos3. nmos3.contact
nmos4.poisson nmos4.electron nmos4. nmos4.contact
4.7 [Example 4.2] Simulation of SRAM of 3D FinFET with Lg = 15 nm 207
System{
Vsource_psetvdd (dd 0) { dc = 0.0 }
Vsource_psetvwl (T 0) { dc = 0.0 }
Vsource_psetvb (L 0) { dc = 0.0 }
Vsource_psetvbl (R 0) { dc = 0.0 }
Vsource_pset vin (VinL 0) { dc = 0.0 }
NMOS1 nmos1( "S"=0 "D"=VinR "G"=VinL )
NMOS2 nmos2( "S"=0 "D"=VinL "G"=VinR )
PMOS1 pmos1( "S"=dd "D"=VinR "G"=VinL )
PMOS2 pmos2( "S"=dd "D"=VinL "G"=VinR )
NMOS3 nmos3( "S"=VinR "D"=L "G"=T )
NMOS4 nmos4( "S"=R "D"=VinL "G"=T )
Plot "n@node@_sys_des.plt" (time() v(VinL) v(dd) v(T) v(L) v(R) v(VinR)
)}
*----------- SRAM circuit scheme -------------------*
setout_filen@previous@_sys_des
proj_load "${out_file}.plt"
cv_createDS inv1 "${out_file} v(VinL)" "${out_file} v(VinR)"
cv_createDS inv2 "${out_file} v(VinR)" "${out_file} v(VinL)"
SRAM of 3D FinFET
Lg=15nm
Vdd =1V
The bigger SNM is the better
Fig. 4.14 Butterfly curve of SRAM based on Lg = 15 nm FinFET. The SNM is 120 mV
nFinFET and pFinFET must be perfectly matched in order to enhance the SNM as
shown in Fig. 4.14, where |Vthp| = |Vthn| and Ion,p = Ion,n, such that the static noise
margin can also be enhanced. With continuous scaling of FinFET, the static noise
margin has become more and more challenging.
In summary, the standard TCAD simulation examples of Inverter and SRAM
with Lg = 15 nm FinFET as the fundamental transistor have been provided in this
chapter. The numeric results of such simulation are in compliance with the current
14-nm/16-nm technology nodes of current semiconductor industry. Readers can
understand the operating mechanisms of Inverter and SRAM based on these two
examples, and they can serve as the reference for continuous scaling.
References
1. C.C. Hu, Modern Semiconductor Devices for Integrated Circuits (PERSON, 2010)
2. Y. Taur, T.H. Ning, Fundamentals of Modern VLSI Device, 2nd edn. (Cambridge University
Press, New York, 2010)
3. C.H. Jan, F. Al-amoody, H.Y. Chang, T. Chang, Y.W. Chen, N. Dias, W. Hafez, D. Ingerly,
M. Jang, E. Karl, S.K.Y. Shi, K. Komeyli, H. Kilambi, A. Kumar, K. Byon, C.G. Lee, J. Lee,
T. Leo, P.C. Liu, N. Nidhi, R. Olac-vaw, C. Petersburg, K. Phoa, C. Prasad, C. Quincy, R.
Ramaswamy, T. Rana, L. Rockford, A. Subramaniam, C. Tsai, P. Vandervoorn, L. Yang, A.
Zainuddin, P. Bai, A 14 nm SoC platform technology featuring 2nd generation tri-gate
transistors, 70 nm gate pitch, 52 nm metal pitch, and 0.0499 lm2 SRAM cells, optimized for
low power, high performance and high density SoC products. VLSI Tech. Symp. T12 (2015)
Chapter 5
Gate-All-Around (GAA) NWFET
with Lg = 10 nm Simulation
According to MOSFET scaling rule, the depletion layer formed in the channel of
traditional 2D MOSFET near source and drain, the short-channel effect (SCE) has
become inevitable along with the scaling of Lg dimension. As an important device
design parameter, k (nature length) relates to SCE, and it is depended to the geo-
metric structure of device as shown in Fig. 5.1. Small k value indicates that the
channel is less vulnerable to the effect of depletion region in source and/or drain
with biasing. Therefore, various multigate device structures have proposed in the
industry for small k value to reduce SCE, such as double-gate FET, tri-gate FET,
FinFET. One of the superior solutions is gate-all-around nanowire FET (GAA
NWFET or GAA FET) structure [1–6]. The evolution of MOSFET from planer,
FinFET to GAA, is shown in Fig. 5.2. GAA FET could be the optimal solution
based on previous description due to the smallest k0.
IBM introduced CMOS logic device and circuit performance of Si
gate-all-around (GAA) nanowire MOSFET (NWFET) in 2013 [3], in which men-
tioned that the development of high performance CMOS logic technological
application in the future would be heading toward GAA due to its excellent gate
control capability, high Ion/Ioff ratio, and extremely high density. Figure 5.3 shows
3D stacking GAA NWFET structure. It has three advantages: (1). Multiple NWs
stacked based on the 3D stacking technology can lead to higher ON current, (2).
Superior gate control capability can reduce SCE. (3). It is compatible to current
CMOS FinFET technology. In short, GAA FET has superior performance and good
candidate for next-generation technology. In Fig. 5.4, 2015 ITRS version 2.0
predicts GAA will apply after the year 2024.
Following section, we will discuss the design guideline and simulation of 3D
GAA NWFET.
ε si
Single gate λ1 = tsi tox
ε ox
ε si
Double gate λ2 = tsi tox
2ε ox
ε si
Trigate (FinFET) λ3 = tsi tox
3ε ox
ε si
Quadruple gate λ4 ≅ tsi tox
4ε ox
2tox
Surrounding gate 2ε si t 2 si ln(1+ ) + ε ox t 2 si
t si
(GAA) λ0 =
16ε ox
Gate
Gate Gate
Bulk-Si SOI
Planar
UTB Planar FinFET
Gate
Tri-Gate GAA
3D-NWFET
Si
Gate
Si
Si
High-κ
BOX
Fig. 5.4 Selected logic core device technology road map as predicted by 2015 ITRS version 2.0
[1, 2]
214 5 Gate-All-Around (GAA) NWFET with Lg = 10 nm Simulation
The following three main program code files are based on Synopsys
Sentaurus TCAD 2014 version. For the simulation tools detailed content of SDE !
devise_dvs.cmd, SDEVICE ! dessis_des.cmd, and INSPECT ! inspect_inc.
cmd, are identical to Example 3.1.
In this section, we only introduce the SDE tool codes (Figs. 5.5 and 5.6).
L g =10nm
Vd =0.7V
Vg =0.7V
Fig. 5.6 The required simulation tools shown in the workbench of n-type GAA NWFET
simulation
5.2 [Example 5.1] 3D IM n-Type GAA NWFET 215
1. SDE ! devise_dvs.cmd
The electrical property Id–Vg of simulation result is shown in Figs. 5.7 and 5.8,
with important parameters of
1. SS at around 72 mV/dec.,
2. Vtn at around 0.32 V,
3. Isat at around 7.2 10−6 A, and
Vd = 0.7 V
Vd = 0.05 V
Drain Current, I d (A)
3D – Electron concentration
L g =10nm
Vd =0.7V
Vg =0.7V
2D – Electron concentration
L g = 10 nm
V g = 0.7 V
V d = 0.7 V
3D-Electric-Field
L g =10nm
Vd =0.7V
Vg =0.7V
2D-Electric-Field
L g =10nm
Vd =0.7V
Vg =0.7V
3D – Electrostatic Potential
L g =10nm
Vd =0.7V
Vg =0.7V
3D – Hole concentration
L g =10nm
Vd =0.7V
Vg =0.7V
Lg = 10 nm, Vd = 0.7, 0.05 V and Vg = 0.7 V, respectively. The reader can follow
the following steps to analyze the physical property of 3D n-type GAA NWFET.
The electron concentration distribution of 3D n-type GAA NWFET simulation
in Fig. 5.9 can be converted into 2D by cutting along X-axis, which is the Y-Z cross
section, as shown in Fig. 5.10 as the channel cross section.
222 5 Gate-All-Around (GAA) NWFET with Lg = 10 nm Simulation
2D – Hole concentration
L g =10nm
Vd =0.7V
Vg =0.7V
L g =10nm
Vd =0.7V
Vg =0.7V
Fig. 5.16 Electron current density distribution of 3D n-type GAA NWFET simulation
The following three main program code files are based on Synopsys
Sentaurus TCAD 2014 version. For the simulation tools detailed content of
SDE ! devise_dvs.cmd, SDEVICE ! dessis_des.cmd, and
INSPECT ! inspect_inc.cmd. Please refer to Examples 3.1 and 5.1.
In this section, we can introduce the SDE tool codes in doping section.
5.3 [Example 5.2] 3D IM p-Type GAA NWFET 223
Band Diagram
Channel
Energy(eV)
L g =10nm
Vd =0.7V
Vg =0.7V
Channel Direction,X(μm)
Fig. 5.18 The required simulation tools are shown in the workbench of p-type GAA NWFET
simulation
The Id–Vg curve of simulation result is shown in Fig. 5.18 with important
parameters of
1. SS at around 69 mV/dec.,
2. Vtp at around −0.32 V,
3. Isat at around 5.35 10−6 A, and
4. Ioff at around 4.93 10−12 A.
The results are from Fig. 5.19 with conditions of Lg = 10 nm, Vd = −0.7,
−0.05 V, and Vg = −0.7 V (Fig. 5.20).
The next is the discussion of physical property analysis of 3D p-type
GAA NWFET. The structural channel mesh, the electron concentration distribu-
tions of 3D and 2D structures, electrical field distributions, electrical potential
distributions, and the energy band diagrams along the channel direction are shown
GAA pNWFET Lg = 10 nm
Vd = -0.7 V
Drain Current, Id (A)
Vd = -0.05 V
L g =10nm
Vd =-0.7V
Vg =-0.7V
L g =10nm
Vd =-0.7V
Vg =-0.7V
Fig. 5.22 Mesh diagram (including SiO2) of 3D p-type GAA NWFET simulation
226 5 Gate-All-Around (GAA) NWFET with Lg = 10 nm Simulation
3D – Hole concentration
L g =10nm
Vd =-0.7V
Vg =-0.7V
2D – Hole concentration
L g =10nm
Vd =-0.7V
Vg =-0.7V
in Figs. 5.21, 5.22, 5.23, 5.24, 5.25, 5.26, 5.27, 5.28, 5.29, 5.30, and 5.31 with the
conditions of Lg = 10 nm, Vd = −0.7, −0.05 V, and Vg = −0.7 V, respectively.
The reader can follow the following steps to analyze the physical property of 3D
p-type GAA NWFET.
5.4 [Example 5.3] 3D Cylindrical IM n-Type GAA NWFET 227
3D – Electron concentration
L g =10nm
Vd =-0.7V
Vg =-0.7V
2D – Electron concentration
L g =10nm
Vd =-0.7V
Vg =-0.7V
L g =10nm
Vd =-0.7V
Vg =-0.7V
Fig. 5.27 Hole current density distribution of 3D p-type GAA NWFET simulation
L g =10nm
Vd =-0.7V
Vg =-0.7V
Fig. 5.28 Hole current density distribution of 2D p-type GAA NWFET simulation
5.4 [Example 5.3] 3D Cylindrical IM n-Type GAA NWFET 229
3D-Electric-Field
L g =10nm
Vd =-0.7V
Vg =-0.7V
2D-Electric-Field
L g =10nm
Vd =-0.7V
Vg =-0.7V
Band Diagram
L g =10nm
Channel
V d =-0.7V
V g =-0.7V
Energy(eV)
Channel Direction,X(μm)
Fig. 5.32 Simulation device parameters of 3D cylindrical IM n-type GAA NWFET with different
radius r = 3, 4, and 5 nm and HfO2 = 2 nm
5.4 [Example 5.3] 3D Cylindrical IM n-Type GAA NWFET 231
D
Lg=10nm
G
r=5nm S
Lg=10nm D
Ch
HfO2
Fig. 5.36 Electron current density of 2D Y–X channel cross section of 3D cylindrical IM n-type
GAA NWFET with radius (r) of 3 nm
5.4 [Example 5.3] 3D Cylindrical IM n-Type GAA NWFET 233
Id(A)
r=4nm, SS~65
Vg (V)
References
1. J.P. Colinge, Multiple-gate SOI MOSFETs. Solid-State Electron. 48, 875 (2004)
2. N. Singh, K.D. Buddharaju, S.K. Manhas, A. Agarwal, S.C. Rustagi, G.Q. Lo, N.
Balasubramanian, D.L. Kwong, Si, SiGe nanowire devices by top–down technology and their
applications. IEEE Trans. Electron Devices 55, 3107 (2008)
3. K. Nayak, M. Bajaj, A. Konar, P.J. Oldiges, K. Natori, H. Iwai, K.V.R.M. Murali, V.R. Rao,
CMOS logic device and circuit performance of si gate all around nanowire MOSFET. IEEE
Trans. Electron Devices 61, 3066 (2014)
4. ITRS version 2.0 (2015), http://www.semiconductors.org/main/2015_international_
technology_roadmap_for_semiconductors_itrs/
5. M.S. Yeh, Y.J. Lee, M.F. Hung, K.C. Liu, Y.C. Wu, High-performance Gate-all-around
poly-Si thin-film transistors by microwave annealing with NH3 plasma passivation. IEEE
Trans. Nanotechnol. 12, 636 (2013)
6. H.B. Chen, C.Y. Chang, N.H. Lu, J.J. Wu, M.H. Han, Y.C. Cheng, Y.C. Wu, Characteristics of
Gate-all-around junctionless poly-Si TFTs with an ultrathin channel. IEEE Electron Device
Lett. 34, 897 (2013)
Chapter 6
Junctionless FET with Lg = 10 nm
Simulation
6.1 Foreword
The diagrams of JL—FET and IM-FET are as shown in Fig. 6.1a, b. As for the
traditional IM-FET, with voltage applied to drain, the junction of drain and channel
is at reverse bias and the depletion region is widening, thus causing the effective
gate length (Lg) to be reduced. This is the so-called SCE. In addition, drain voltage
will also reduce the energy barrier of channel, which is known as drain-induced
barrier lowing (DIBL). These two aforementioned phenomena are not obvious in
the MOSFET with long channel. However, along with the scaling of channel
length, these two phenomena will cause Vth to be reduced, which is known as Vth
roll-off based on the equation as shown below:
LGate
(a) Vd
TSi D
G
S
Vs TSi
(b)
JNT
Vs n+ n+ n+ Vd
IM
Vs n+ p n+ Vd
Fig. 6.1 a Aerial views of JL—FET (JNT) and IM-FET (IM), b longitudinal concentration
distribution of JNT and IM [5]
6.3 JL—FET Operating Mechanism 239
(a) (c)
Gate Gate
JL—FET N+
N+ N+ IM-FET N+
off-state P-type
Lphysical Lphysical
Leff Leff
LSCE
(b)
Gate
JL—FET
N+ on-state N+
Lphysical
Leff
Fig. 6.2 Comparison among different Leff. a The Leff with the channel of JL—FET in off-state.
b The Leff with the channel of JL—FET in on-state. c The existence of junction when the channel
of IM-FET in off-state will lead to smaller LSCE, thus making it vulnerable to SCE
By compared to IM-FET, the difference is the dopant type and dopant concentration
in the active layer as shown in Fig. 6.2. The source, drain, and channel of JL—FET
are all based on the same doping and concentration. The comparison between
different operating mechanisms of (a) inversion-mode (IM) ‘‘N+PN+’’ FET,
(b) accumulation-mode (AC) ‘‘N+NN+’’ FET, (c) junctionless-mode (JL)
‘‘N+N+N+’’ FET are as shown in Fig. 6.3.
(a) In n-type IM-FET, the depletion region will take place between flatband voltage
(Vfb) and Vth. If it is greater than Vth, there will be a strong inversion on the
surface of channel thus forming an inversion layer.
Vfb
Log(Id )
Log(Id )
Log(Id )
Vfb
Vth Vth Vth
IM AC JL
Vfb
Vgs Vgs Vgs
Fig. 6.3 Drain current as the function of Vg, a inversion-mode (IM), b accumulation-mode (AC),
c junctionless-mode (JL) [7]
240 6 Junctionless FET with Lg = 10 nm Simulation
(b) In n-type AC-FET, it is fully depleted when Vg is lower than Vth. When Vg is
between Vth and Vfb, the channel will be partially depleted. When Vg continues
to rise, an accumulation layer will be formed on the surface of channel.
(c) In n-type JL—FET, source, drain and channel are all heavily doped around
1019–1020 cm−3. The channel is fully depleted when Vg is lower than Vth. The
electron concentration in the channel will be increased along with increasing
Vg. The channel concentration will approach maximum concentration when Vg
reaches Vth with dopant concentration of Nd. When Vg = Vfb > Vth, the channel
concentration is identical to source and drain concentration.
The electron concentration cross sections of n-type JL—FET are as shown in
Fig. 6.4. The diagram of fixed Vd = 50 mV versus different Vg are as shown in
Fig. 6.4a–d. These phenomena are obtained from the results of simulations based
on Poisson Equation and the drift-diffusion model (Fig. 6.4).
(a) When Vg is lower than Vth, the n-type channel will be fully depleted, and the
device is in OFF state.
(b) When Vg is equaled to Vth, the electron concentration of n-type silicon channel
gradually becomes close to the dopant concentration while connecting drain
and source, and free electrons are gradually generated in the channel along with
the reduction of depletion region.
(c) When Vg is higher than Vth, the electron concentration of n-type silicon channel
has become identical to the dopant concentration.
(d) When Vg reaches Vfb, the electron concentration at the junction of n-type silicon
channel dopant concentration is identical to source and drain concentration. The
JL—FET current conduction behavior can be regarded as a resistor.
The drain current of JL—FET is different from that of IM-FET. The former is
more like a resistor relying on the drift current of majority carriers, and the latter
relies on the drift current of minority inverse carriers. The equation of drain satu-
ration current of general long-channel IM-FET (Idsat) is shown as:
(a) (b)
Drain Drain
Source Source
(c) (d)
Drain Drain
Source Source
Fig. 6.4 Electron concentration cross sections of n-type JL—FET with fixed Vd = 50 mV.
a Vg < Vth; b Vg = Vth; c Vg > Vth; d Vg = Vfb Vth [7]
6.3 JL—FET Operating Mechanism 241
W
Idsat lCox ðVg Vth Þ2 ð6:2Þ
L
where W is the channel width, L is the gate length, Vg is the gate voltage, and Cox is
the capacitor of oxide layer. On the other hand, JL—FET is a normally on device
just like a resistor. When Vg is equaled to Vfb, the Idsat:
Tsi Wsi
Idsat qlNd Vg ð6:3Þ
L
where Tsi is the thickness of silicon channel, and Nd is the dopant concentration. It
is shown in this equation that Idsat is not related to Cox.
The current conduction positions of IM-FET, AC-FET, and JL—FET are as
shown in Fig. 6.5. In IM-FET when it is lower than Vth, the current conduction
position is at the corner because of greater electric field here as shown in Fig. 6.5a.
When it is higher than Vth, the current takes place in the channel along the edge of
gate as shown in Fig. 6.5d; in AC-FET when it is lower than Vth, the sub-threshold
current is formed in the center of channel as shown in Fig. 6.5b. When it is higher
than Vth, it will result in the surface current similar to inversion-mode as shown in
Fig. 6.5e; in the end, in JL—FET when it is lower than Vth, the sub-threshold
current is also formed in the center of channel just like AC-FET as shown in
Fig. 6.5c. When it is higher than Vth, the channel is partially depleted, and the main
current is still contributed by the center of channel as shown in Fig. 6.5f. The
conducting current is concentrated in the center, which is called body current. It is
different to the conventional MOSFET as surface current underneath of gate oxide.
The advantage of this body current of JL—FET can reduce the interface scattering
effects in the channel.
Design and simulation of 3D JL—FET by using Synopsys Sentaurus TCAD
2014 version will be described in this chapter.
distributions of
inversion-mode, 44 44 44
accumulation-mode, and
junctionless-mode FETs. The
FETs in the state of “Below
Threshold” are as shown in
(a) BOX (b) BOX (c) BOX
the upper half, and the FETs
in the state of “Above
Threshold” are as shown in 44 44 44
the lower half
The Id–Vg curve of simulation result is shown in Fig. 6.7 with important
parameters as shown in Fig. 6.8. From Figs. 6.7 and 6.8, it appears that the SS of
Lg = 10 nm p-type JL—FET is around 73 mV/dec. with excellent switch charac-
teristic. The structural channel mesh, the electron concentration distributions of 3D
and 2D structures, electric field distributions, electric potential distributions, and the
energy band diagrams along the channel direction are as shown in Figs. 6.9, 6.10,
6.11, 6.12, 6.13, 6.14, 6.15, 6.16 and 6.17 with the conditions of Lg = 10 nm,
Vd = 1 V, and Vg = 1 V, respectively. The additional texts in Figs. 6.9, 6.10, 6.11,
6.12, 6.13, 6.14, 6.15, 6.16 and 6.17 are added via PowerPoint for better under-
standing by readers.
Fig. 6.6 Required simulation tools are shown in the workbench for n-type JL—FET
244 6 Junctionless FET with Lg = 10 nm Simulation
3D-nJLFET
Vd = 1 V
Vd = 0.05 V
Drain Current, Id (A)
S.S. = 73 mV/dec @ Lg = 10 nm
Fig. 6.7 Id–Vg curve of 3D simulation of Lg = 10 nm n-type JL—FET (some descriptions are
completed by PowerPoint after snapshot by Inspect)
3D-nJLFET Mesh
Lg =10 nm
Vd =1V
Vg =1V
3D-nJLFET Mesh
Lg =10 nm
Vd =1V
Vg =1V
3D-Electron concentration
Lg =10 nm
Vd =1V
Vg =1V
2D-Electron concentration
Lg =10 nm
Vd =1V
Vg =1V
3D-Electric Field
Lg =10 nm
Vd =1V
Vg =1V
1. SDE – devise_dvs.cmd
This is the standard example of 3D JL—FET.
The line of code following; is the prompt character for program designer to
take note such that it will not be executed by the computer.
6.5 [Example 6.2] p-Type JL—FET with Lg = 10 nm 247
2D-Electric Field
Lg =10 nm
Vd =1V
Vg =1V
3D-Electrostic Potential
Lg =10 nm
Vd =1V
Vg =1V
2D-Electrostic Potential
Lg =10 nm
Vd =1V
Vg =1V
Band Diagram
Channel
Energy (eV)
Lg =10 nm
Vd =1V
Vg =1V
Fig. 6.17 Energy band diagram along the channel direction of simulation of 3D Lg = 10 nm
n-type JL—FET in on-state
6.5 [Example 6.2] p-Type JL—FET with Lg = 10 nm 249
Fig. 6.18 Required simulation tools are shown in the workbench for p-type JL—FET
The Id–Vg curve of simulation result is as shown in Fig. 6.19 with important
parameters as shown in Fig. 6.20. From Figs. 6.19 and 6.20, it appears that the SS
of Lg = 10 nm p-type JL—FET is around 73 mV/dec. with good electric properties.
The structural channel mesh, the electron concentration distributions of 3D and
2D structures, electric field distributions, electric potential distributions, and the
energy band diagrams along the channel direction are as shown in Figs. 6.21, 6.22,
6.23, 6.24, 6.25, 6.26, 6.27, 6.28 and 6.29 with the conditions of Lg = 10 nm,
Vd = −1 V, and Vg = −1 V, respectively. The additional texts in Figs. 6.21, 6.22,
6.23, 6.24, 6.25, 6.26, 6.27, 6.28 and 6.29 are added via PowerPoint for better
understanding by readers.
250 6 Junctionless FET with Lg = 10 nm Simulation
3D-pJLFET
Vd = -1 V
Vd = -0.05 V
Drain Current, Id (A)
3D-pJLFET Mesh
Lg =10 nm
Vd =-1V
Vg =-1V
3D-nJLFET Mesh
Lg =10 nm
Vd =-1V
Vg =-1V
3D-Hole concentration
Lg =10 nm
Vd =-1V
Vg =-1V
2D-Hole concentration
3D-Electric Field
Lg =10 nm
Vd =-1V
Vg =-1V
2D-Electric Field
Lg =10 nm
Vd =-1V
Vg =-1V
3D-Electrostic Potential
Lg =10 nm
Vd =-1V
Vg =-1V
2D-Electrostic Potential
Lg =10 nm
Vd =-1V
Vg =-1V
Band Diagram
Lg =10 nm
Channel
Vd =-1V
Vg =-1V
Energy (eV)
Fig. 6.29 Energy band diagram along the channel direction of simulation of 3D Lg = 10 nm
p-type JL—FET in on-state
6.5 [Example 6.2] p-Type JL—FET with Lg = 10 nm 255
References
1. H.B. Chen, Y.C. Wu, C.Y. Chang, M.H. Han, N.H. Lu, Y.C. Cheng, Performance of GAA
poly-Si nanosheet (2 nm) channel of junctionless transistors with ideal subthreshold slope.
VLSI Technology Symposium, p T232 (2013)
2. Y.C. Cheng, H.B. Chen, C.S. Shao, J.J. Su, Y.C. Wu, C.Y. Chang, T.C. Chang, Performance
enhancement of a novel P-type junctionless transistor using a hybrid Poly-Si fin channel.
Technical Digest of IEDM, 26.27.21 (2014)
3. M.S. Yeh, Y.C. Wu, M.H. Wu, Y.R. Jhan, M.H. Chung, M.F. Hung, High performance
ultra-thin body (2.4 nm) poly-Si junctionless thin film transistors with a trench structure.
Technical Digest of IEDM, 26.26.21 (2014)
4. S. Migita, Y. Morita, M. Masahara, H. Ota, Electrical performances of junctionless-FETs at the
scaling limit (Lch = 3 nm). Technical Digest of IEDM, 8.6.1 (2012)
5. J.P. Colinge, I. Ferain, G. Fagas, S. Das, P. Razavi, R. Ya, Influence of channel material
properties on performance of nanowire transistors. J. Appl. Phys. 111, 124509 (2012)
6. J.P. Colinge, I. Ferain, A. Kranti, C.W. Lee, N.D. Akhavan, P. Razavi, R. Ya, R. Yu,
Junctionless nanowire transistor: complementary metal-oxide-semiconductor without junc-
tions. Sci. Adv. Mater. 3, 477 (2011)
7. J.P. Colinge, C.W. Lee, N. Dehdashti Akhavan, R. Yan, I. Ferain, P. Razavi, A. Kranti, R. Yu,
Semiconductor-On-Insulator Materials for Nanoelectronics Applications. Engineering
Materials, Springer, Berlin (2011)
Chapter 7
Steep Slope Tunnel FET Simulation
The simplest way for increasing the transistor density in the wafer is to reduce the
feature size of transistor. During scaling down of feature size by Moore’s law, the
supply voltage Vdd must also be reduced in accordance with the principle of
constant electrical field scaling rule. And the threshold voltage Vth must also be
reduced along with the reduction of Vdd in order to maintain the consistent
overdrive voltage (Vov = Vdd − Vth) and Ion as shown in Fig. 7.1.
For the fixed Vdd = 1 V, by reduction of Vth, the high-performance (high Vov and
high Id) device will also lead to high leakage current (Ioff). On the other hand, by
increasing Vth, the Ioff reduce will lead to low Id and device performance. The
phenomenon comes from the subthreshold slope (SS) > 60 mV/dec. (SS = kT/
q ln10 ~60 mV/dec.) and will result in significant increase of off-state current
level. The correlations among Vov, SS, Vth, Ioff, and Ion are as shown in Fig. 7.1.
The calculation of IC energy consumption is very important, especially when the
requirement of current portable products, which need low power consumption. The
total consumed energy can be divided into dynamic energy consumption (Edynamic)
and static energy consumption (Eleakage), and the relationship equation between total
consumed energy (J) and Vdd is as shown below [1]:
Ioff 1E-9
SS
Ioff 1E-11
Vg
1
f ¼ ð7:2Þ
Ld sdelay
CVdd
sdelay ¼ ð7:3Þ
Ion
It is clearly shown in Eq. (7.1) and Fig. 7.1 that the transistor with low standby
power consumption can be achieved by working on small SS and low Vdd. Vdd will
depend on the operating voltage Vov = (Vdd −Vth), and the excessive reduction of
this voltage will result in another issue of low Ion. Therefore, we can reduce the
off-state current Ioff by another method of reducing SS value.
It is shown in Fig. 7.1 that Ioff is limited by Vth and subthreshold slope (SS). The SS
physical limit is 60 mV/dec. If SS < 60 mV/dec., Ioff can be greatly reduced in
Fig. 7.2. The SS of conventional MOSFET is resulted from thermal diffusion current,
such that it is bound to be >60 mV/dec [1]. The current of tunnel FET (TFET) is
resulted from tunneling current rather than thermal diffusion current, such that the SS
of TFET can break the physical limit of 60 mV/dec. In recent years, there have been
many international research groups dedicated to study on innovative TFET structures
and materials to break SS of 60 mV/dec, and also increasing the ON current [2–6].
Figure 7.3 shows TFET is a perfect choice for ultra-low-power device with low
Vth and low Vdd, which is urgently needed for Internet of Thing (IoT)
ultra-low-power applications as shown in Fig. 7.4.
7.2 Operating Mechanism of Tunnel FET (TFET) 259
I d ∝ (V dd -V th)2
10-7A
A: SS>60mV/dec
B: SS=60mV/dec
C: SS<60mV/dec
Vg
V th
SS=60mV/dec
Vg
Vth= 0.1V 0.3V
The fundamental structure of TFET is as shown in Fig. 7.5, and the basic
operating mechanism of TFET based on the principle of “Gate-controlled reverse
PIN diode”, which is shown in Fig. 7.6. Unlike conventional n-channel MOSFET,
the drain of n-channel TFET is based on n-type doping, and the source is based on
p-type doping. For n-channel TFET, the electrons tunnel from p-type doping source
to n-type doping drain. The TFET operation is shown in Fig. 7.6.
In the off-state, electrons in drain and holes in source will both face rather high
energy barrier thus preventing conduction of both electrons and holes, and the
leakage current Ioff is also fairly low. In the on-state, the gate voltage is bigger than
threshold voltage (Vth), and a significant band bending will occur in the depletion
region in the area of source close to the channel, and the width of depletion region will
be gradually reduced along with the increasing gate voltage (Vg). Under such
260 7 Steep Slope Tunnel FET Simulation
Fig. 7.4 Coming era of IoT and the applications of various semiconductor devices requiring
ultra-low-voltage operations
P+ N+
Source Drain
Intrinsic Channel
conditions, electrons in the valence band of source will have the chance to tunnel
through the narrow depletion region (*nm) and enter the conduction band of
channel, and this is the unique tunneling current conduction mechanism of TFET.
The conduction current is called tunneling current which is proportional to the tun-
neling probability of WKB approximation is based on the equation as shown below:
pffiffiffiffiffiffiffiffi 3=2 !
4k 2m Eg
Ion / Twkb exp ð7:4Þ
3qhðEg þ DUÞ
- Vd = Vdd = 1 V
on-state
Vg = 1 V Vth = 0.3 V
- Vd = Vdd = 1 V
of valence band (Ev) of source and the energy of conduction band (Ec) of channel
(Fig. 7.5).
The conduction current of TFET is also related to the difference between the
values of Fermi-Dirac distribution functions of channel and source and tunneling
probability as shown in the equation below:
ð7:5Þ
where Dch and Ds are the density of states of channel and source, respectively, and
fch(E) and fs(E) are the Fermi-Dirac functions of channel and source.
The following 3D nTFET three main program code files are based on Synopsys
Sentaurus TCAD 2014 version.
In the introduction of TFET fundamental structure in Fig. 7.5, the channel is
intrinsic without any doping. However, the channel in Example 7.1 is based on a
slight amount of n-type doping for the purpose of adjusting threshold voltage (Vth).
262 7 Steep Slope Tunnel FET Simulation
1. SDE—devise_dvs.cmd
The method for establishing SDE of 3D nTFET is similar to Chap. 3 such that
the code will not be introduced in details here. For detailed information please
refer to Chap. 3. Here, we only introduce the SDE tool codes: devise_dvs.cmd,
especially note the doping section.
"ABA"
;--- Source contact and Source ---;
(sdegeo:create-cuboid (position 0 0 0 ) (position x1 y1 z1 ) "Silicon" "SourceC")
(sdegeo:create-cuboid (position x1 0 0 ) (position x2 y1 z1) "Silicon" "Source")
;--- Gate oxide ---;
(sdegeo:create-cuboid (position x2 ( - Tox) 0 ) (position x3 y2 z2 ) "SiO2"
"Gateoxide")
;--- Channel ---;
(sdegeo:create-cuboid (position x2 0 0 ) (position x3 y1 z1 ) "Silicon" "Channel")
;--- Drain contact and Drain---;
(sdegeo:create-cuboid (position x3 0 0 ) (position x4 y1 z1 ) "Silicon" "Drain")
(sdegeo:create-cuboid (position x4 0 0 ) (position x5 y1 z1 ) "Silicon" "DrainC")
;--- Buried oxide ---;
(sdegeo:create-cuboid (position 0 (- 10) (- 10) ) (position x5 y3 0 ) "SiO2" "Box")
;----------------------------------------- Contact ----------------------------------------;
;----- Source -----;
(sdegeo:define-contact-set "S" 4.0 (color:rgb 1.0 0.0 0.0 ) "##" )
(sdegeo:set-current-contact-set "S")
(sdegeo:set-contact-faces (find-face-id (position 1 1 z1)))
;----- Drain -----;
(sdegeo:define-contact-set "D" 4.0 (color:rgb 1.0 0.0 0.0 ) "##" )
(sdegeo:set-current-contact-set "D")
(sdegeo:set-contact-faces (find-face-id (position (+ x4 1) 1 z1 )))
;----- Front Gate -----;
(sdegeo:define-contact-set "G" 4.0 (color:rgb 1.0 0.0 0.0 ) "||" )
(sdegeo:set-current-contact-set "G")
(sdegeo:set-contact-faces (find-face-id (position (+ x2 1) (- Tox) 1 )))
;----- Top Gate -----;
(sdegeo:define-contact-set "G" 4.0 (color:rgb 1.0 0.0 0.0 ) "||" )
(sdegeo:set-current-contact-set "G")
(sdegeo:set-contact-faces (find-face-id (position (+ x2 1) 1 z2 )))
;----- Back Gate -----;
(sdegeo:define-contact-set "G" 4.0 (color:rgb 1.0 0.0 0.0 ) "||" )
(sdegeo:set-current-contact-set "G")
(sdegeo:set-contact-faces (find-face-id (position (+ x2 1) y2 1 )))
;----------------------------------------- Doping -----------------------------------------;
264 7 Steep Slope Tunnel FET Simulation
Fig. 7.7 Required simulation tools are shown in the workbench for nTFET
2. SDEVICE—dessis_des.cmd
The method for establishing SDEVICE of 3D nTFET is similar to s in Chap. 3
such that the code will not be introduced in details here. For detailed information
please refer to Chap. 3. The current of TFET is tunneling current, such that the
physical model of tunneling Band2Band(E2) must be added in dessis_des.
cmd. For the physical details please refer to SDEVICE manual of
Sentaurus TCAD 2014 version.
...............................
Physics{
Mobility( DopingDep HighFieldsat Enormal )
EffectiveIntrinsicDensity( OldSlotboom BandGapNarrowing
(BennettWilson ) )
Recombination( SRH(DopingDependence) Auger Band2Band(E2) )
*ComputeIonizationIntegrals(WriteAll)
eQuantumPotential
hQuantumPotential }
...........................
3. INSPECT—inspect_inc.cmd
The method for establishing INSPECT of 3D nTFET is the same as in Chap. 3
such that the code will not be introduced in details here. For detailed information
please refer to Chap. 3.
The Id–Vg curve of simulation result is as shown in Fig. 7.8, and the important
parameters are as shown in Fig. 7.9. Sometime, the point hopping occurs to IV
characteristic in the subthreshold region; the SS extracted by Inspect can no longer
serve as the reference. For obtaining precise SS, readers can obtain the text output
266 7 Steep Slope Tunnel FET Simulation
3D-nTFET
Vd = 0.7 V
Vd = 0.5 V
Drain Current, I d (A)
Vd = 0.1 V
SS = 50 mV/dec.
Lg = 100 nm
3D-nTFET Mesh
Lg = 100 nm
Vd = 0.7 V
Vg = 3 V
WK = 4.5 eV
Fig. 7.10 Mesh diagram of the simulation of 3D n-type TFET (gated reverse PIN diode)
7.3 Example 7.1 (Design and Simulation of 3D n-Type TFET) 267
3D-Electron concentration
Lg = 100 nm
Vd = 0.7 V
Vg = 3 V
WK = 4.5 eV
2D-Electron concentration
g = 100 nm
LG
VDd = 0.7 V
VGg= 3 V
WK = 4.5 eV
3D-Electric Field
Lg = 100 nm
Vd = 0.7 V
Vg = 3 V
WK = 4.5 eV
2D-Electric Field
Lg = 100 nm
Vd = 0.7 V
Vg = 3 V
WK = 4.5 eV
The following three main program code files are based on Synopsys
Sentaurus TCAD 2014 version. The drain doping concentration D_Doping is set
as variable.
TFET is the transistor which can be subjected to bipolar operation, which means
it can be turned on by either positive or negative bias. This phenomenon will cause
7.4 Example 7.2 (3D n-Type TFET of Different Drain Doping Concentrations) 269
3D-Electrostatic Potential
Lg = 100 nm
Vd = 0.7 V
Vg = 3 V
WK = 4.5 eV
2D-Electrostatic Potential
Lg = 100 nm
Vd = 0.7 V
Vg = 3 V
WK = 4.5 eV
the large off-state current (Ioff) and it cannot be used for IC. This problem can be
solved by reducing the dopant concentration of drain. The impact of reduced drain
concentration of n-type TFET demonstrates in Example 7.2.
1. SDE—devise_dvs.cmd
The method for establishing SDE of 3D nTFET is similar to Chap. 3. The only
special part is about the doping. Therefore, only the program code of doping is
introduced here. For the rest of codes please refer to Chap. 3.
270 7 Steep Slope Tunnel FET Simulation
Band Diagram
Lg = 100 nm
Source
Vd = 0.7 V
Vg = 3 V
Energy (eV) e
Channel Drain
Fig. 7.17 Energy band diagram along the channel direction of simulation of 3D n-type TFET in
on-state. The VB of source overlaps to CB of channel, and electrons will tunnel from source to
drain through channel
Fig. 7.18 Required simulation tools are shown in the workbench for nTFET with different drain
doping concentrations
7.4 Example 7.2 (3D n-Type TFET of Different Drain Doping Concentrations) 271
It is indicated in the code that source is doped by boron and Drain is doped by
arsenic. Unlike the aforementioned introduction, the arsenic of lower concentration
is used as the dopant for the channel for the purpose of adjusting Vth
2. SDEVICE—dessis_des.cmd
The method for establishing SDEVICE of 3D nTFET is the same as in Chap. 3,
so the codes will not be introduced in details here. Readers can refer to Chap. 3
for detailed information.
3. INSPECT—inspect_inc.cmd
The code of this part is completely identical to Example 7.1, so readers can
directly refer to the code of Example 7.1.
The Id–Vg curve of simulation result of Example 7.2 is as shown in Fig. 7.19,
with important parameters as shown in Fig. 7.20. It is indicated in Fig. 7.19 that the
bipolarity of TFET can be effectively inhibited by reducing the dopant concentra-
tion of drain. In this simulation results, the Ioff can be reduced by decreasing dopant
concentration of drain. The drain doping 1E18 cm−3 shows lowest Ioff than others.
It can be explained that the lower drain doping has less energy band bending to
prevent the leakage current.
272 7 Steep Slope Tunnel FET Simulation
Doping = 1e20
Drain Current, Id (A)
SS = 50 mV/dec.
L g = 100 nm
Vd = 0.7 V
Vg = 3 V
WK = 4.5 eV
Doping = 1e18
Fig. 7.19 Id–Vg curves of simulation of 3D n-type TFET with different drain dopant
concentrations
Fig. 7.20 Electric property parameters of simulation of 3D n-type TFET with different drain
dopant concentrations
current (2.47 10−5 A/lm), because the screening length (k) of a GAA nanowire
(NW) structure is half that of the planar structure. Simulations reveal that a sub-
threshold swing (SS) as low as 42 mV/dec, and an on/off current ratio as higher as
1010 is realized. The AG-TFET is easily fabricated as an actual device by simply
changing the layout of gate in a general TFET fabrication.
1. SDE—devise_dvs.cmd
The method for establishing SDE of AG-TFET is the same as in Chap. 3, so the
codes will not be introduced in details here. Readers can refer to Chap. 3 for
detailed information.
2. SDEVICE—dessis_des.cmd
The method for establishing SDEVICE of AG-TFET is the same as in Chap. 3,
so the codes will not be introduced in details here. Readers can refer to Chap. 3
for detailed information. The current of TFET is tunneling current, so the tun-
neling physical model must be added in the physics part with the manual as the
reference for details.
3. INSPECT—inspect_inc.cmd
The method for establishing INSPECT of AG-TFET is the same as in Chap. 3,
so the codes will not be introduced in details here. Readers can refer to Chap. 3
for detailed information.
Figure 7.21 displays the architecture of the AG-TFET and the parameters of its
simulation. The total channel length is 20 nm. The left half (10 nm) of gate of
AG-TFET is controlled by the surrounding gate with a square cross section, and the
X = 50 (nm)
Lg
X = 30
20 nm N+
Drain
Gate
10 nm
5 nm
10 nm
P+ 15 nm Device Structure AG-TFET
Source
Gate Length 20 nm
GAA cross-section area 5 5 nm2
Fig. 7.21 Device structure and important simulation parameters of n-type AG-TFET [7]
274 7 Steep Slope Tunnel FET Simulation
right half (10 nm) is controlled by the planar gate. The cross-sectional area of the
GAA channel is 5 5 nm2 and that of the planar channel is 15 20 nm2. The
effective oxide thickness is 1.3 nm (to meet the International Technology Roadmap
for Semiconductors: ITRS), and the gate work function is 4.72 eV (to meet TiN).
The doping concentrations of the p-type source, the n-type drain, and the low-doped
n-type channel are 1 1020, 1 1019, and 1 1016 cm−3, respectively.
Figure 7.22 compares the transfer characteristics of the n-channel AG-TFET to
those of the gate-all-around (GAA) and the planar TFET. The simulated on-state
current (Ion) in a GAA TFET is 2.11 10−5 A/lm at Vg = 2 V, and the off-state
current (Ioff) in a planar TFET is 1.51 10−15 A/lm at Vg = 0 V. The SS of the
GAA TFET and the planar TFET is 61 mV/dec and 124 mV/dec, respectively.
Therefore, the AG-TFET combines the advantages of both structures, with a
2679-fold higher Ion than that of the planar TFET and a 476-fold lower Ioff than that
of the GAA TFET. When this asymmetric-gate architecture is used in the TFET, the
minimum SS is 42 mV/dec, the average SS is 45 mV/dec (determined over three
decades of Id), and the maximum Ion/Ioff ratio is 1010.
Figures 7.23 and 7.24 present simulated energy band diagrams of the AG-TFET
in the on-state (Vg = 2 V) and the off-state (Vg = −0.5 V), respectively. The energy
band of the right half-channel of the AG-TFET will be effectively shifted down as
the full GAA channel in the on-state. When the bias voltage is large enough to
reduce the barrier width, electrons tunnel from the valence band of the source side
to the conduction band of the channel side. In a TFET, the triangular barrier width
is the screening tunneling length (k). The screening lengths of a GAA and a planar
structure, k1 and the k2, respectively, are given by
10 -4 30x10 -6
GAA
10 -5 Planar Vd = 0.5 V
10 -6 AG-TFET 25x10 -6
2679
Drain Current (A/μm)
10 -7 AG-TFET
GAA 20x10 -6
10 -8
Planar AG-TFET
10 -9 15x10 -6
SS min = 42 mV/dec
10 -10 SS avg = 45 mV/dec
10 -11 Ion /Ioff ≈ 10 10 10x10 -6
10 -12
10 -13 5x10 -6
10 -14
476
0
10 -15
10 -16
-0.5 0.0 0.5 1.0 1.5 2.0
Gate Voltage ( V )
Energy (eV)
direction in on-state 0.0
(Vg = 2 V, Vd = 0.5 V) [7] - -
-0.5
λ1
-1.0 Vg = 2 V Drain
-1.5 Vd = 0.5 V
-2.0
Channel direction, X(nm)
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
eSi
k1 ¼ TSi Tox ð7:6Þ
4eox
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
eSi
k2 ¼ TSi Tox ð7:7Þ
eox
where esi and eox are the dielectric constants of silicon and oxide, respectively. Tsi
and Tox are the thickness of the silicon and the oxide, respectively. The value of k1
is half that of k2. Accordingly, the tunneling probability in the GAA structure is
higher than in the planar structure. Figure 7.23 reveals that using the GAA structure
on the source side increases the Ion of a TFET. When the TFET is operated in the
off-state, electrons tunnel from the valance band of the channel side to the con-
duction band of the drain side, producing a leakage current in the TFET. Thus,
Fig. 7.24 indicates that the planar structure that is used at the drain side reduces the
Ioff of the TFET owing to its large screening tunneling length.
Figures 7.25 and 7.26 present the BTBT generation rates in the channel direction
at the center of the channel in the AG-TFET at Vg = 2 V and Vg = −0.5 V,
respectively. The source tunneling junction is formed at X = 30 nm (NW region),
and the drain tunneling junction is generated at X = 50 nm (planar region). As
expected, the BTBT generation rate is highest on the shortest tunneling path, as
presented in Figs. 7.23 and 7.24. Therefore, the generation rate in the planar
structure is six orders of magnitude smaller than that in the GAA structure.
276 7 Steep Slope Tunnel FET Simulation
-3S-1)
generation in channel Vd = 0.5 V
direction in n-type AG-TFET
at Vg = 2 V, Vd = 0.5 V [7] 40x10 30
20x10 30
BTBT Generat
10x10 30
60 70
W
idt 55 60 m)
hd 50 X (n
ire 50 40 n ,
cti 30 ectio
on 45 20 ir
,Y d
(nm 40 10 nnel
) a
Ch
at Vg = −0.5 V, Vd = 0.5 V
[7] 60x10 24
50x10 24
40x10 24
30x10 24
BTBT Generat
20x10 24
10x10 24
60 70
W 60 )
idt 55 50 nm
h d 50 40 , X(
ire 30 on
cti 45 cti
on
,Y
20
d ire
40 10 l
(nm ne
an
) Ch
Figure 7.25 shows BTBT generation rate peaks close to the gate dielectric,
as predicted by the screening length formulas 7.6 and 7.7. The peak in Fig. 7.26 is
at the center of the channel, because the electric field is concentrated there.
The channel series resistance of the AG-TFET is lower than that of the
GAA TFET, so the AG-TFET has a higher on-current than does the GAA TFET.
Figure 7.27 compares the output characteristics of the AG-TFET, the GAA TFET,
and the planar TFET. Clearly, the AG-TFET has a higher saturation drain current
than the GAA TFET because the AG-TFET has a lower series resistance. The
AG-TFET has a lower off-current and SS, because its planar part has a larger
screening length. The planar TFET has the largest screening length, and therefore
the lowest drain current. Figure 7.28 plots the output characteristic of the AG-TFET
7.5 Example 7.3 (3D n-Type TFET with Asymmetrical Gate) 277
40
Vg = 2V
AG-TFET
20 GAA
Planar
0
0.0 0.5 1.0 1.5 2.0
Drain Voltage (V)
Vg = 1.2 V
6
4 Vg = 1.1 V
Vg = 1.0 V
2
as a function of the gate voltage. Its low parasitic resistance and excellent drain
current saturation behavior reveal its potential for use in future low-power inte-
grated circuits.
278 7 Steep Slope Tunnel FET Simulation
This chapter demonstrated the standard example of TFET simulation. The electrical
properties indicate SS < 60 mV/dec., the ultra-low leakage current, Ioff (in fA), and
extremely high Ion/Ioff ratio. The aforementioned simulation results show that TFET
is very suitable for future application of ultra-low-power semiconductor device.
References
8.1 Foreword
Huge efforts are put into CMOS scaling to push the limits of Moore’s law.
Semiconductor ICs manufacturing companies are currently ramping up
16-nm/14-nm FinFET processes, with 7 and 5 nm technology nodes just around the
corner. As we approach sub-10-nm node technologies, different device models have
been proposed and intensively researched to overcome the several critical challenges
that arise due to the relentless scaling to ever small dimensions. Various approaches
have been proposed and comprehensively explored to attenuate the impact of
short-channel effect (SCE) on threshold voltage, drain-induced barrier lowering
(DIBL), and subthreshold swing (SS). Leakage current (Ioff) and electrostatics (gate
control) become important factors of concern. High-k dielectrics and high mobility
materials and various device architectures are explored extensively. There are a few
papers that successfully address challenges in ultra-scaled gate length real devices
[1, 2].
Finding a suitable semiconductor material for the sub-10-nm technology is the
major challenge for the semiconductor researchers around the world. The materials
that are investigated need to be compatible to the current CMOS technology
adapted by the industry. The next-generation materials that possess similar qualities
as that of Silicon which is cost efficient as well as that suits the industry requirement
is very difficult to identify. The reliability of the newly investigated materials meets
many challenges. The silicon and germanium technologies are more mature and are
most researched for many decades now. Suitable continuous scaling of transistors
hinders the development of high-quality junctions especially in sub-10-nm nodes
where modifying the doping concentration becomes an even more strenuous
process.
(a) Silicon
Silicon is one of the most researched materials worldwide. It is abundance in nature,
easy to handle, robustness, and cheaper cost made it a favorite candidature in
semiconductor industry. For the past two decades, scientists around the world
successfully devised many new ways to scale down Si-based devices. Hence, the Si
technology is extremely mature and more reliable than any other semiconductor
material.
(b) Germanium
The first transistor emerged from germanium almost seven decades ago. But it
represents very small market today because of its instability with a lower bandgap
energy compared to Si. On the other hand, Ge has higher electron and hole
mobility. Thus, Ge devices can function at higher frequencies than Si devices. This
makes Ge a promising candidate for sub-10-nm node. Also, Ge device technology
is similar to that of current industrial Si technologies.
(c) III-V and 2D high mobility semiconductor materials
Other important materials are mainly III-V compound semiconductors composed
of elements of group III (basically Al, Ga, and In) and elements of group V
(basically N, P, As, and Sb). Among a total of 12 combinations, the combinations
most likely to replace silicon include GaAs, InP, GaP, GaN, and InAs. Recently,
there have been many new 2D materials being studied such as graphene and MoS2.
However, these higher mobility materials still face numerous problems including
mass automotive production challenge, threshold voltage (Vth) control of nFET and
pFET challenge, high off-sate leakage current (Ioff), reliability challenge, and mass
production cost challenge.
8.1 Foreword 281
In this section, we propose the design guideline of sub-20-nm to 9-nm gate length
(Lg) Si FinFET wine-bottle shape Fin structure. The real sub-20-nm experimental
results show that taller and thinner FinFET has lower Ioff and lower drain-induced
barrier lowering (DIBL). Through TCAD simulation, we predict that the wine-bottle
shape FinFET is promising future sub-10-nm FinFET with excellent gate control to
eliminate short-channel effect (SCE) and sufficient volume for epitaxing low resis-
tance raised source and drain (S/D) materials. The Vth and SS are all reasonable and
insensitive (DVth < 14 mV, DSS < 3 mV/dec) to various sizes of wine-bottle shape
Fin structure. The Ion increases with Fin height (Fh), top Fin weight (Fwt) increases
monotonically, and Ioff is vice verse. The wine-bottle FinFET is promising for
Moore’s law that can extend to sub-10-nm Lg CMOS IC technology.
G
Top width (Fwt ) Fh L g(nm) 28nm 20nm 12nm 9nm
B Fwb(nm) 12 12 5 5
o
d
Y S/D Dop. 8E19 8E19 8E19 8E19
W= 20nm
H= 80nm Ch. Dop. 8E18 8E18 5E18 5E18
Fig. 8.1 Device structure and parameters of simulated wine-bottle FinFET, with Fin height (Fh),
top Fin width (Fwt), and bottom Fin width (Fwb)
Figure 8.4 shows simulated linear Ids–Vgs sub-20-nm FinFET plots versus (a) Fh
and (b) Fwt. The Ids is highly depending on Fwt rather than Fh. The ion increases
with the Fwt increasing.
Figure 8.5 shows simulated linear Ids–Vgs curve of wine-bottle FinFET keep the
trapezoidal Fin of (a) pFET and (b) nFET. The results reveal that the Ion can be
increased (+18%) by using tall Fh and wide top Fin width (Fwt).
Figure 8.6 shows simulated 3D contour plot influence of Fh and Fwt for
sub-20-nm wine-bottle nFinET and pFinET with Vth, Ion, and Ioff, respectively. The
Vth and SS (not shown) are all reasonable and insensitive values (Vth < 14 mV,
SS < 3 mV/dec). The ion increases with Fin height (Fh), top Fin weight (Fwt)
increases monotonically, and Ioff is in opposite trend. Once achieving the target Vth
and Ioff values, the Ion can be increased by using larger Fwt.
Figure 8.7 shows simulated 3D contour plot influence of Fh and Fw for
Lg = 20-nm wine-bottle nFinET and pFinET with Vth, Ion, and Ioff, respectively. The
Vth and SS are all reasonable and insensitive values (DVth < 8 mV,
DSS < 4 mV/dec). The Vth insensitivity reveals that the Vtn and Vtp can entirely
adjust by using proper metal gate materials with different work functions. The Ion
increases with Fin height (Fh), top Fin weight (Fwt) increases monotonically, and
Ioff is in opposite trend.
Figure 8.8a plots the simulated timing characteristics of a Si CMOS inverter
circuit of simulated Lg = 12-nm wine-bottle FinFETs. The Thl is 0.89 ps, and Tlh is
1.8 ps. Figure 8.8b plots simulated SRAM characteristics with signal noise margin
(SNM) of 160 mV.
8.3 Study of Silicon Lg = 3 nm Bulk IM, AC … 283
nFinFET
L g =28nm
pFinFET
L g =28nm
Fig. 8.2 Simulation results of electron and hole density at off-state of (a) sub-20 nm n-type and
(b) p-type FinFET. The tall FinFET has lower electron and hole density distribution than STD
FinFET
-4
(a) 10-4 (b) 10 Vd =-0.7V Vd =0.7V
-5 -5
10 Vd =-0.8V Vd =0.8V 10
-6 -6
10 10
-7 -7
10 10
Ids (A)
-8 -8 Vd =-0.05V Vd =0.05V
Ids (A)
10 Vd =-0.05V Vd =0.05V 10
SS~68 ~67
-9
SS~64 SS~62 -9
10 10
-10 -10
10 10
-11 L g=28nm -11
10 Fh=50nm 10 Fh=50nm L g=20nm
Fwt=6nm EOT=0.5nm -12 Fwt=6nm EOT=0.5nm
-12
10 Fwb=12nm Vg =0.8V 10 Vg =0.7V
Fwb=12nm
-13 -13
10 10
-0.8 -0.4 0.0 0.4 0.8 -0.8 -0.4 0.0 0.4 0.8
Vgs (V) Vgs (V)
Vd =-0.05V
Ids (A)
-8 -8
10 SS~84 SS~68 10 SS~86 SS~75
-9 -9
10 10
-10 -10
10 10
10
-11 Fh=50nm Lg=12nm 10
-11
Fh=50nm Lg=9nm
Fwt=3nm EOT=0.3nm Fwt=3nm EOT=0.3nm
-12 -12
10 Fwb=5nm Vg=0.6V 10 Fwb=5nm Vg=0.5V
-13 -13
10 10
-0.8 -0.4 0.0 0.4 0.8 -0.8 -0.4 0.0 0.4 0.8
Vgs (V) Vgs(V)
Fig. 8.3 Simulated Ids–Vgs of (a) Lg = 28 nm, (b) Lg = 20 nm, (c) Lg = 12 nm, and
(d) Lg = 9 nm wine-bottle FinFET
node will be completely different from that of the devices in higher technology
nodes. The sub-10-nm technology devices will be more strictly adhering to the laws
of quantum physics and important quantum confinement phenomenon, and
size-dependent properties will come to effect more severely in sub-10-nm node.
Hence, it is important to compare and analyze the performance of the conventional
inversion-mode of operation along with the other modes of operation such as
accumulation-mode and junctionless-mode. We examine the performance of the
optimized 3-nm FinFET with homogeneous source and drain doping concentration
in inversion-mode (IM), accumulation-mode (AC), and junctionless-mode
(JL) operation. The transfer and output characteristics in IM, AC, and JL modes
of simulated sub-5-nm technology node devices are discussed in detail. In addition,
8.3 Study of Silicon Lg = 3 nm Bulk IM, AC … 285
(a) (b)
5x10-6 6x10-6
Fh=60nm
Fh=58nm Fwt =6nm
4x10-6 Fh=56nm 5x10-6 Fwt =5nm
Fh=54nm Fwt =4nm
Fh=52nm
Fh=50nm
4x10-6 Fwt =3nm
3x10-6 Fwt =2nm
Fh=48nm
Ids (A)
Ids (A)
Fh=46nm 3x10-6
2x10-6 Vd=-0.8V Vd=0.8V
Vd=-0.8V Vd=0.8V
2x10-6
1x10-6
Fwt=4nm 1x10-6 Fh =50nm
Fwb=12nm Fwb=12nm
0 0
-0.8 -0.4 0.0 0.4 0.8 -0.8 -0.4 0.0 0.4 0.8
Vgs (V) Vgs (V)
Fig. 8.4 Linear Ids–Vgs sub-20-nm FinFET plots versus (a) Fh and (b) Fwt. The Ids is highly
increasing with the Fwt
(a) 5x10
-6
(b) 6x10-6
-6
-6 5x10
4x10
-6
-6
4x10
Ids (A)
3x10
Ids (A)
-6
3x10
-6
2x10 -6
2x10
-6 -6
1x10 Top/Bot Fw= 6/14 nm 1x10 Top/Bot Fw=6/14 nm
Top/Bot Fw= 4/12 nm Top/Bot Fw=4/12 nm
Top/Bot Fw= 2/10 nm Top/Bot Fw=2/10 nm
0 0
-0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 0.3 0.4 0.5 0.6 0.7 0.8
Vgs (V) Vgs (V)
Fig. 8.5 Linear Ids–Vgs sub-20-nm FinFET trapezoidal Fin of (a) pFET and (b) nFET
for each case, we interpret the 3D electron density mesh plots. The device per-
formances such as the drain-induced barrier lowering, subthreshold slope, and
on/off current ratio have also been estimated. This chapter serves as only a design
guideline and in future with more ab initio and first principle-based models can be
incorporated in the device physics for more accurate results.
In this section, we investigated the device performance of the optimized 3-nm
gate length (Lg) bulk silicon FinFET device using 3D quantum transport device
simulation. By keeping source and drain doping constant and by varying only the
channel doping, the simulated device is made to operate in three different modes
such as inversion-mode (IM), accumulation-mode (AC), and junctionless-mode
(JL). The excellent electrical characteristics of the 3-nm gate length Si-based bulk
FinFET device were investigated. The subthreshold slope values
(SS * 65 mV/dec) and drain-induced barrier lowering (DIBL < 17 mV/V) are
analyzed in all three IM, AC, and JL modes of bulk FinFET with |Vth| * 0.31 V.
286 8 Extremely Scaled Si and Ge to Lg = 3 nm FinFETs …
Fig. 8.6 (a–c) are 3D plots of Lg = 28-nm nFET with Vth, Ion, and Ioff. (d–f) are 3D plots of pFET
with Vth, Ion, and Ioff, respectively. The Fwb is fixed at 12 nm
Fig. 8.7 (a–c) are 3D plots of Lg = 20-nm nFinET with Vth, Ion, and Ioff, respectively. (d–f) are
3D plots of pFET with Vth, Ion, and Ioff, respectively. Fwb is fixed at 12 nm
Furthermore, the threshold voltage (Vth) of the bulk FinFET can be easily tuned by
varying the work function (WK). This research reveals that Moore’s law can
continue up to 3-nm nodes.
The simulated device structure and the table of important parameters used in the
device simulation are given in Fig. 8.9. We applied equivalent oxide thickness
(EOT) of 0.3 nm. The gate length (Lg) is 3 nm, and the Fin width (Fw) and the Fin
8.3 Study of Silicon Lg = 3 nm Bulk IM, AC … 287
0.8 0.5
(V)
Voltage (V)
out1 out 2
0.4 0.3
/V
0.2 THL=0.89ps TLH=1.8ps 0.2
0.0 0.1
V
-0.2 0.0
0 2e-11 4e-11 6e-11 8e-11 0.0 0.1 0.2 0.3 0.4 0.5 0.6
Time (s) Vout2 / Vin1 ( V )
Fig. 8.8 Simulated Lg = 12-nm FinFET (a) inverter timing characteristics and (b) SRAM
characteristics with signal noise margin (SNM) of 160 mV
height (Fh) are also the same (Fw = Fh = 3 nm). The doping concentrations of
source/drain in all three modes (IM, AC, JL) of bulk FinFET devices are set to
1.0 1020 cm−3 for both n-type and p-type transistors. The channel concentration
of JL bulk FinFET is set to 1.0 1020 cm−3. The channel concentration of IM and
AC bulk FinFET is set to 1.0 1018 cm−3. Arsenic and boron are used as dopants
in device simulation. The bulk doping concentration for the FinFET is
5 1018 cm−3, which can be implemented easily by usual well doping implanta-
tion. A constant Vth value was maintained for both nFET (Vth * 0.31 V) and pFET
(Vth * −0.31 V) in all three modes of operation. The work function used for
n-type IM, AC, and JL modes is 4.40, 4.41, and 4.55 eV, respectively. Similarly,
the work function for p-type IM, AC, and JL modes is 4.80, 4.81, and 4.69 eV,
respectively. Precise numerical results of the simulated nanoscale device are
obtained by solving 3D quantum transport equations provided by Synopsys
Sentaurus version 2014. In quantum transport equations, a density-gradient model
is used in the simulation. The bandgap narrowing model and Shockley–Read–Hall
recombination with doping-dependent model are also considered. The mobility
model used in device simulation is according to Matthiessen’s rule.
np n2i;eff
RSRH ¼ ð8:1Þ
sp ðn þ n1 Þ þ sn ðp þ p1 Þ
EF;p Ev Kp
p ¼ Nv F1=2 ð Þ ð8:2Þ
kTp
288 8 Extremely Scaled Si and Ge to Lg = 3 nm FinFETs …
Lg = 3nm
EOT = 0.3nm
h
w
Z
X
STI
Y
Substrate Doping N : 5x1018 cm-3 ,P-Type N : 5x1018 cm-3 ,P-Type N : 5x1018 cm-3 ,P-Type
Concentration P : 5x1018 cm-3 ,N-Type P : 5x1018 cm-3 ,N-Type P : 5x1018 cm-3 ,N-Type
Fig. 8.9 Device structure and important parameters of simulated 3-nm gate length (Lg) IM, AC,
and JL Si bulk FinFET [3]
EF;n Ec Kn
n ¼ Nc F1=2 ð Þ ð8:3Þ
kT n
mp and mn are effective mass of hole and electron, respectively, and Etrap is the
difference between defect energy level and intrinsic energy level. In addition, the
mobility model in the device simulation is in accordance with the following
Matthiessen’s rule:
8.3 Study of Silicon Lg = 3 nm Bulk IM, AC … 289
1 D D 1
¼ þ þ ð8:6Þ
l lsurf aps lsurf rs lbulk dop
In D = exp(x/lcrit), x is the distance from the interface, and lcrit is the fitting
parameter. The mobility is composed of three kinds of phenomena, such as acoustic
phonon scattering (lsurf_aps), surface roughness scattering (lsurf_rs), and bulk
mobility with doping-dependent modification (lbulk_dop).
Results and Discussion
Left-hand side plots of Figs. 8.10, 8.11, and 8.12 show the Id–Vg curves of the
n-type and p-type devices of interest, in which the linear threshold voltage (Vth) is
all adjusted to approximately ±300 mV for fair comparison. In the proposed n-type
IM, AC, and JL bulk FinFET, the saturation current (at Vg = 0.7 V, Vd = 1 V) is
2.52 10−4 A/µm, 2.54 10−4 A/µm, and 2.32 10−4 A/µm, respectively. For
p-type IM, AC, and JL bulk FinFET, the saturation current is 2.24 10−4,
2.25 10−4, and 2.26 10−4 A/µm, respectively. The SS for n-type IM, AC, and
JL modes is, respectively, 78.74, 78.79, and 77.37 mV/dec. The SS for p-type IM,
AC, and JL modes is 67.63, 67.65, and 62.28 mV/dec, respectively. The DIBL,
defined as the difference in Vth between Vd = 0.05 V and Vd = 0. 7 V, for n-type
IM, AC, and JL modes, equals only 16.04, 16.17, and 26.80 mV/V, respectively.
The similar performances are also achieved in p-type IM, AC, and JL bulk
FinFET (29.80, 31.89 mV/V, and 40.20 mV/V). The DIBL and SS numerical values
are tabulated in Table 8.1. As the Fw and Fh are reduced to 3 nm, the SS and DIBL
approach to their ideal value (60 mV/dec and 0 mV/V) in the simulated results.
Si nFET
0.5
-3
10 Vds =-0.7V Si pFET Vds = 0.7V Si nFET
IVg Vth I= 0.4 V to 0.8 V
Step = 0.2 V Si pFET
-4
Drain Current (mA/µ m)
10 0.4
Drain Current (A/ µ m)
JL Mode
-5 JL Mode Vds =0.05V
10 Vds =-0.05V Lg =3nm
Lg =3nm
0.3
-6
10
-7
10 Si pFET SS : Si nFET SS : 0.2
62.28 mV/dec 77.37 mV/dec
-8
10
Si pFET DIBL : Si nFET DIBL : 0.1
-9 40.20 mv/V 26.80mv/V
10
-10
10 0.0
-1.0 -0.5 0.0 0.5 1.0 -2 -1 0 1 2
Gate Voltage (V) Drain Voltage (V)
Fig. 8.10 Id–Vg of 3-nm gate length (Lg) for n-type and p-type Si bulk FinFET operating in JL
mode with SS and DIBL values shown inset and Id–Vd of 3-nm gate length (Lg) for n-type and
p-type Si bulk FinFET operating in JL mode, with overdrive voltage |Vov| = |Vg − Vth| [3]
290 8 Extremely Scaled Si and Ge to Lg = 3 nm FinFETs …
0.5
-3 Si nFET
10 Vds =-0.7V Si nFET Vds = 0.7V IVg Vth I= 0.4 V to 0.8 V Si pFET
Si pFET Step = 0.2 V
0.4
AC Mode
-5 AC Mode Vds =0.05V L g =3nm
10 Vds =-0.05V
Lg =3nm 0.3
-6
10
-7
10 Si pFET SS : Si nFET SS :
0.2
67.65 mV/dec 78.79 mV/dec
-8
10
Si pFET DIBL : Si nFET DIBL : 0.1
-9 31.89mv/V 16.17 mv/V
10
-10
10 0.0
-1.0 -0.5 0.0 0.5 1.0 -2 -1 0 1 2
Gate Voltage (V) Drain Voltage (V)
Fig. 8.11 Id–Vg of 3-nm gate length (Lg) for n-type and p-type Si bulk FinFET operating in AC
mode with SS and DIBL values shown inset and Id–Vd of 3-nm gate length (Lg) for n-type and
p-type Si bulk FinFET operating in AC mode, with overdrive voltage |Vov| = |Vg − Vth| [3]
0.5
-3 Si nFET Vds = 0.7V IVg Vth I= 0.4 V to 0.8 V Si nFET
10 Vds =-0.7V Si pFET Si pFET
Step = 0.2 V
0.4
Drain Current (mA/µm)
-4
10
Drain Current (A/µ m)
IM Mode
-5 IM Mode Vds =0.05V
10 Vds =-0.05V Lg =3nm
Lg =3nm 0.3
-6
10
-7
10 Si pFET SS : Si nFET SS : 0.2
67.63 mV/dec 78.74 mV/dec
-8
10
Si pFET DIBL : Si nFET DIBL : 0.1
-9 29.80mv/V 16.04 mv/V
10
-10
10 0.0
-1.0 -0.5 0.0 0.5 1.0 -2 -1 0 1 2
Gate Voltage (V) Drain Voltage (V)
Fig. 8.12 Id–Vg of 3-nm gate length (Lg) for n-type and p-type Si bulk FinFET operating in IM
with SS and DIBL values shown inset and Id–Vd of 3-nm gate length (Lg) for n-type and p-type Si
bulk FinFET operating in IM, with overdrive voltage |Vov| = |Vg − Vth| [3]
It is noteworthy that the off-state current is all low in IM, AC, and JL modes of
Si bulk FinFET owing to extensively scaled nanofin.
Right-hand side plots of Figs. 8.10, 8.11, and 8.12 show the output characteristic
curves of Si FinFET. It is very clear that simulated IM and AC devices have almost
similar Id–Vd output characteristic curves.
Figure 8.13a, b compares on-state (Vgs = 1 V) and off-state (Vgs = 1 mV)
electron density distribution at the 3D cross sections of the 3-nm nanofin n-type Si
bulk JL-FinFET. The conduction path is located at the middle of the nanofin as
expected. Figure 8.13c–f compares on-state (Vgs = 1 V) and off-state (Vgs = 1 mV)
8.3 Study of Silicon Lg = 3 nm Bulk IM, AC … 291
Table 8.1 Important numerical values of simulated 3-nm gate length IM, AC, and JL Si bulk
FinFETs [3]
Device mode Junctionless Accumulation Inversion
Work function (eV) N: 4.55 N: 4.40 N: 4.40
P: 4.60 P: 4.80 P: 4.80
Vth (*0.31 V) N: 0.3057 N: 0.2987 N: 0.3000
P: −0.2969 P: −0.2987 P: −0.3019
SS (mV/dec) N: 77.37 N: 78.79 N: 78.74
P: 62.28 P: 67.65 P: 67.63
DIBL (mv/V) N: 26.80 N: 16.17 N: 16.04
P: 40.20 P: 31.89 P: 29.80
Ion (A/lm) N: 2.32 10−4 N: 2.54 10−4 N: 2.522 10−4
P: 2.26 10−4 P: 2.24 10−4 P: 2.240 10−4
electron density distribution at the 3D cross sections of the 3-nm nanofin n-type Si
bulk IM-FinFET and AC-FinFET, respectively. Notably, the on-state and off-state
results of 3-D eDensity distribution from quantum transport simulation demonstrate
that the device can be scaled down to a physical limit of 3-nm node. The current
conduction in all three (JL, AC, and IM) modes is almost similar because the
carriers fully occupy 3-nm nanofin cross section. The electrons are more concen-
trated at the middle topside of channel in IM, AC, and JL bulk FinFET as better
controllability by gate is achieved with Lg = Fw = Fh = 3 nm.
In summary, we have performed various analyses in the 3-nm gate length bulk
silicon FinFET operating in inversion-mode (IM), accumulation-mode (AC), and
junctionless-mode (JL). The observed transfer characteristics, output characteristics,
and electron density distribution results of the 3D quantum transport device sim-
ulation reveal the fact that all the three IM, AC, and JL modes of operation are
perfectly feasible even at 3-nm gate length. Thus, it enables the bulk FinFET
devices to be scaled down to its least possible physical limits obeying Moore’s
scaling law.
In this section, the Synopsys Sentaurus TCAD 2014 version 3D device simulation
is used to show the performances of n-type and p-type 3-nm bulk Ge FinFET of
IM-FET, AC-FET, and JL—FET. The simulated bulk Ge FinFET device exhibits
better short-channel characteristics, including drain-induced barrier lowering
(DIBL < 10 mV/V) and subthreshold slope (SS * 64 mV/dec). Electron density
distributions in on-state and off-state also show that the simulated devices have
better Ion/ Ioff ratios.
292 8 Extremely Scaled Si and Ge to Lg = 3 nm FinFETs …
(#/cm ) (a) Z
(b)
Y
3
(#/cm )
3
1e+20 1e+15 G
X
8e+19 8e+14
G
Electron Density
ity
6e+19 6e+14
Electron Dens
4e+19 4e+14
3.0 3.0
2e+19 2.5 2e+14 2.5
)
m
2.0 2.0
)
nm
(n
0 1.5 1.5
0
2.5 1.0
2.5 1.0
ht
t(
2.0
1.5 0.5 2.0 0.5
1.5
ig
h
1.0
Fin W 0.0 1.0
ig
0.5
Fin W 0.0
He
id 0.5
id
He
(Subs th (nm) (S u b th (n m )
n
trate) n s t r a te
Fi
Fi
)
(c) (d)
G G
(#/cm )
3
1e+20 2e+14
(#/cm )
3
2e+14
8e+19 G 2e+14 G
ity
1e+14
6e+19
ity
Electron Dens
1e+14
Electron Dens
1e+14
4e+19
8e+13
2e+19 3.0 6e+13
)
2.5
m
3.0
2.0 4e+13
(n
2.5
m)
0 1.5 2e+13 2.0
ht
2.5 1.0
(n
0 1.5
ig
2.0 0.5
Fin W 1.5 1.0
ht
2.5
He
1.0
0.0 2.0
id
eig
( S u b th ( n m )
0.5
Fin W 1.5 1.0 0.5
n
s tr a t id
Fi
nH
0.0
(S u bs t h ( n m )
0.5
e)
t r at e )
(e) Fi
(f)
ity (#/cm )
G
3
1.4e+20
G
(#/cm )
3
1.8e+14
1.2e+20
1.0e+20 G 1.6e+14
1.4e+14 G
ity
8.0e+19
Electron Dens
1.2e+14
Electron Dens
6.0e+19 1.0e+14
8.0e+13
4.0e+19
3.0 6.0e+13
2.0e+19 2.5
)
3.0
m
2.0 4.0e+13
2.5
(n
2.0
2.5 1.0
ht
(n
2.5 1.0
ht
id id
(Sub th (nm)
0.0
S u b s th (n m )
He
0.5
n
Fi
strat trate
e)
n
Fi
Fig. 8.13 3D mesh plot for electron density distributions in the 3-nm gate length (Lg) n-type Si
bulk FinFET in (a) JL on-state (Vgs = 1 V) and (b) JL off-state (Vgs = 1 mV); (c) IM on-state and
(d) IM off-state; and (e) AC on-state and (f) AC off-state [3]
Simulation Method
The simulated device structure and the table of important parameters used in the
device simulation are given in Fig. 8.14. We applied equivalent oxide thickness
(EOT) of 0.3 nm. The gate length (Lg) is 3 nm, and the Fin width (Fw) and the Fin
height (Fh) are also the same (Fw = Fh = 3 nm). The doping concentrations of
8.4 Study of Germanium Lg = 3-nm Bulk FinFET 293
source/drain in all three modes (IM, AC, JL) of bulk FinFET devices are set to
1.0 1020 cm−3 for both n-type and p-type transistors. The channel concentration
of bulk JL-FinFET is set to 1.0 1020 cm−3. The channel concentration of IM and
AC bulk FinFET is set to 1.0 1018 cm−3. Arsenic and boron are used as dopants
in device simulation. The bulk doping concentration for the FinFET is
5 1018 cm−3. A constant Vth value was maintained for both nFET
(Vth * 0.31 V) and pFET (Vth * −0.31 V) in all three modes of operation. The
work function used for n-type IM, AC, and JL modes is 4.40, 4.41, and 4.40 eV,
respectively. Similarly, the work function used for p-type IM, AC, and JL modes is
4.37, 4.37, and 4.33 eV, respectively.
Results and Discussion
Id–Vg curves of Si and Ge 3-nm bulk FinFETs are as shown in Figs. 8.15a, 8.16a,
and 8.17a. The Isat of Ge IM, AC, and JL nFET is around 3.0 10−4 A/µm. The
Isat of Ge IM, AC, and JL pFET is around 3.3 10−4 A/µm. The Isat of Si IM, AC,
and JL nFET is around 3.1 10−4 A/µm. The Isat of Si IM, AC, and JL pFET is
around 2.8 10−4 A/µm.
The SS values for Ge IM nFET and Si IM nFET are 64.38 and 78.74 mV/dec.
The DIBL value for Ge IM pFET and Si IM pFET is 5.43 and 29.80 mV/V, which
is *5 times higher for Si compared to Ge. The simulated AC mode of Ge nFET
Lg = 3nm
EOT = 0.3nm
h
w
STI
Substrate Doping N : 5x1018 cm-3 ,P-Type N : 5x1018 cm-3 ,P-Type N : 5x1018 cm-3 ,P-Type
Concentration P : 5x1018 cm-3 ,N-Type P : 5x1018 cm-3 ,N-Type P : 5x1018 cm-3 ,N-Type
Fig. 8.14 Device structure and important parameters of simulated 3-nm gate length (Lg) IM, AC,
and JL germanium bulk FinFET
294 8 Extremely Scaled Si and Ge to Lg = 3 nm FinFETs …
IM Mode Lg =3nm
-5 Vds =0.05V
10 Vds =-0.05V Lg =3nm
0.3
-6 Ge nFET SS :
10 Ge pFET SS :
64.9 mV/dec 64.38 mV/dec
-7 Si pFET SS : Si nFET SS :
10 78.74 mV/dec 0.2
67.63 mV/dec
Ge pFET DIBL : Ge nFET DIBL :
-8 13.40mv/V
10 05.43mv/V
Si nFET DIBL :
Si pFET DIBL : 0.1
-9 29.80mv/V 16.04mv/V
10
-10
10 0.0
-1.0 -0.5 0.0 0.5 1.0 -2 -1 0 1 2
Gate Voltage (V) Drain Voltage (V)
Fig. 8.15 (a) Id–Vg of 3-nm gate length (Lg) for n-type and p-type Ge bulk FinFET operating in
IM with SS and DIBL values shown inset and (b) Id–Vd of 3-nm gate length (Lg) for n-type and
p-type Ge bulk FinFET operating in IM, with overdrive voltage |Vov| = |Vg − Vth|
Si pFET
10
Drain Current (A/ µm)
AC Mode
-5
AC Mode Lg =3nm
Vds =-0.05V Vds =0.05V
10 Lg =3nm
0.3
-6
10 Ge pFET SS : Ge nFET SS :
64.49 mV/dec 64.38 mV/dec
-7 Si pFET SS : Si nFET SS :
10 0.2
67.65 mV/dec 78.79 mV/dec
-8 Ge pFET DIBL : Ge nFET DIBL :
10 05.75mv/V 14.58 mv/V
Si pFET DIBL : Si nFET DIBL : 0.1
-9 31.89mv/V 16.17 mv/V
10
-10
10 0.0
-1.0 -0.5 0.0 0.5 1.0 -2 -1 0 1 2
Gate Voltage (V) Drain Voltage (V)
Fig. 8.16 (a) Id–Vg of 3-nm gate length (Lg) for n-type and p-type Ge bulk FinFET operating in
AC mode with SS and DIBL values shown inset and (b) Id–Vd of 3-nm gate length (Lg) for n-type
and p-type Ge bulk FinFET operating in AC mode, with overdrive voltage |Vov| = |Vg − Vth|
and Si nFET obtains a SS value of 64.38 and 78.79 mV/dec, respectively, and
DIBL values of Ge AC pFET and Si AC pFET are 5.75 and 31.89 mV/V,
respectively.
Id–Vg of Ge JL nFET achieves almost ideal SS value of 65.92 mV/dec, and Si JL
nFET has SS value of 77.37 mV/dec. The DIBL value of Ge JL pFET and Si JL
pFET is 8.38 and 40.20 mV/V, respectively. It is noteworthy that the off-state
8.4 Study of Germanium Lg = 3-nm Bulk FinFET 295
10-10 0.0
-1.0 -0.5 0.0 0.5 1.0 -2 -1 0 1 2
Gate Voltage (V) Drain Voltage (V)
Fig. 8.17 (a) Id–Vg of 3-nm gate length (Lg) for n-type and p-type Ge bulk FinFET operating in
JL mode with SS and DIBL values shown inset and (b) Id–Vd of 3-nm gate length (Lg) for n-type
and p-type Ge bulk FinFET operating in JL mode, with overdrive voltage |Vov| = |Vg − Vth|
current is all low in IM, AC, and JL modes of both bulk FinFETs owing to
extensively scaled nanofin.
Figs. 8.15b and 8.16b show the Id–Vd curves of IM and AC Ge pFET with an
anomalous kink-effect behavior. It could be explained by the decrease in hole
quantum capacitance (Cq) in IM and AC modes which in turn degrades gate to
channel capacitance in 3-nm nanofin owing to the light hole effective mass
(m*p * 0.04m0) of Ge.
On the other hand, Fig. 8.17b shows that JL Ge pFET has almost negligible kink
effect. It is significant to note that the IM, AC, and JL modes of Si pFET show no
such irregularity, in the Id–Vd characteristics. It must be noted that quantum con-
finement effect (QCE) is involved in this case and the principle is yet to be clarified
on further research study.
The comparisons of electron density distributions and electrostatic potentials of
3D cross sections of 3-nm nanofin n-type Ge IM and JL bulk FinFETs are as shown
in Fig. 8.18, where the conduction path is located at the middle of the nanofin. The
electron density distributions of n-type Ge bulk FinFET in on-state (Vgs = 1 V) and
off-state (Vgs = 1 mV) are as shown in Fig. 8.18a, b.
Table 8.2 summarizes the key parameters used in the quantum transport simu-
lation and significant results which demonstrate high-performance Ge bulk FinFET
at ultra-scaled Lg of 3-nm.
In summary, the Lg = 3-nm Ge bulk FinFETs in operations under IM mode, AC
mode, and JL mode are analyzed in this section. Ge FinFET has better electrical
performance compared to Si FinFET. However, smaller effective mass of Ge results
in degradation of quantum capacitance compared to a standard 3-nm gate length Si
296 8 Extremely Scaled Si and Ge to Lg = 3 nm FinFETs …
(a)
Ge IM N ON Ge JL N ON
Ge IM N ON Ge JL N ON
(b)
Ge IM N OFF Ge JL N OFF
Ge IM N OFF Ge JL N OFF
Fig. 8.18 Electron density (top) and electric field (bottom) distributions in the channel when
n-type JL and IM devices operate at (a) on-state (Vgs = 0.7 V) and (b) off-state (Vgs = 1 mV) with
Lg = Fw = Fh = 3 nm and EOT = 0.3 nm
device. The electron density distribution reveals the fact that with an optimized
3-nm nanofin, charge carriers fully occupy the Fin region in all three modes (JL,
AC and IM) of operation. The observed transfer, output characteristics iterates the
fact that even at 3-nm Lg high-performance Ge bulk FinFET is feasible with all
three IM, AC, and JL modes of operation for future sub-5-nm device applications.
8.5 Study of Silicon and Germanium UTB-JL—FET … 297
The next part is the introduction of simulation of ultra-thin body junctionless FET
(UTB-JL—FET) of Si and Ge with Lg = 1 nm and Lg = 3 nm, which is coupled
with the drift-diffusion (DD) and density-gradient (DG) models for finding solu-
tions. The simulation results indicate that the UTB structure is well suited for Si and
Ge. By using the UTB structure, the short-channel device does not have to be in
compliance with the equation of Tch = Lg/3. In addition, the Ge UTB-JL—FET
6T-SRAM has a reasonable static noise margin (SNM) of 149 mV. The circuit
simulation result shows that UTB-JL—FET can be used for the CMOS technology
node of sub-5 nm.
Junctionless field-effect transistor (JL—FET) structure can circumvent afore-
mentioned issues because the channel region of JL—FET has high doping con-
centrations and the same dopant type as source/drain regions. Owing to the special
doping profile, JL—FET has many advantages such as (1) lower thermal budget
which can integrate with high-k/metal gate easier than conventional MOSFETs,
(2) longer effective channel length than conventional MOSFETs, (3) the body
current which can avoid surface scattering, and (4) avoidance of complicated
source/drain engineering. Therefore, JL—FET is a potential candidate for
ultra-short-channel transistor. But JL—FET has turnoff problem due to high doping
concentrations in channel region. To solve this problem, JL—FET needs ultra-thin
body (UTB) structure to reach fully depleted channel region in off-state.
The UTB structure can provide quantum confinement effect in channel region
which will increase energy bandgap, and this large bandgap can suppress leakage
current. Consequently, an empirical rule of Tch = Lg/3 has been used for the defi-
nition of transistor dimension. As transistor features are scaled, the drive current (Id)
is declined. Therefore, a high mobility material is necessary for sub-10-nm tech-
nology node. Germanium (Ge) is a potential candidate owing to its high mobility.
The electron mobility of Ge is two times higher than Si, and the hole mobility of Ge
is four times higher than Si.
We investigate the electrical performance of Si UTB-JL—FET compared to Ge
with Lg = 1 nm and Lg = 3 nm by 3D simulations. The transistors and circuit
298 8 Extremely Scaled Si and Ge to Lg = 3 nm FinFETs …
performances are discussed in detail. The simulation results reveal that Si and Ge JL
—FET with UTB (1 nm) structure can be employed in sub-5-nm CMOS tech-
nology nodes. Furthermore, this UTB structure can be achieved in the future
technology nodes by focused ion beam (FIB) or reactive-ion etching (RIE). Using
atomic layer chemical vapor deposition system (ALD) and chemical-mechanical
polishing (CMP) processes can perform high-k/metal gate in this UTB-JL—FET for
future application.
Simulation Method
Figure 8.19 displays the architecture of the UTB-JL—FET and the parameters used
in the simulation. The Synopsys TCAD simulator was employed to perform 3D
simulations, which included the coupled drift-diffusion (DD), density-gradient
(DG) model, bandgap narrowing, and quantum effects. A Si and a Ge were used in
the simulated channel material. The channel width is 10 nm. Because of quantum
confinement effect, this ultra-short-channel (Lg = 1 nm) device has normally off
characteristics. As UTB is employed, ultra-short-channel device does not need to
follow an empirical rule of Tch = Lg/3, which is often used as a guideline to sup-
press short-channel effect. We have shown conduction band energy (EC) diagrams
of Si UTB-JL—FET with Lg = 1 nm in both off-state and on-state. In off-state, the
UTB structure builds a high EC level at gated region because of quantum mech-
anism bandgap shift. This energy barrier can block electrons which pass through
channel by thermal injection and direct tunneling. In on-state, owing to the absence
of energy barrier, the electrons pass through channel by ballistic transport in
JL—FET.
Results and Discussion
Figure 8.20a, b shows conduction band energy (Ec) diagrams of Si UTB-JL—FET
with Lg = 1 nm in off-state and on-state, respectively. In off-state, the UTB struc-
ture builds a high Ec level at gated region because of quantum mechanism bandgap
shift. This energy barrier can block electrons which pass through channel by
thermal injection and direct tunneling. In on-state, owing to the absence of energy
barrier, the electrons pass through channel by ballistic transport in JL—FET.
Figure 8.21a, b shows the Id–Vg characteristics of UTB-JL—FET with
Lg = 1 nm in Si and Ge channel, respectively. Owing to the ultra-thin channel, this
device has high Ion/Ioff current ratio of 105 at Vg = 1 V. The SS is 100 mV/decade
of Si pFET and 99 mV/decade of Si nFET, respectively. The DIBL is 225 mV/V of
Si pFET and 222 mV/V of Si nFET, respectively. The SS is 96 mV/decade of Ge
pFET and 93 mV/decade of Ge nFET, respectively. The DIBL is 196 mV/V of Ge
pFET and 200 mV/V of Ge nFET, respectively. Even though this
ultra-short-channel device does not follow an empirical rule of Tch = Lg/3, the
electrical properties can meet the industry requirements because of quantum con-
fined UTB structure. The saturation current is 1.29 10−3 and 1.5 10−3 A/lm
of Si nFET and Ge nFET, respectively. The saturation current is 0.82 10−3 A/lm
and 1.08 10−3 A/lm of Si pFET and Ge pFET, respectively. The Ge nFET has a
16% higher saturation current (Isat) than Si nFET, and the Ge pFET has a 32%
higher Isat than Si pFET.
Figure 8.22a, b shows the Id–Vg characteristics of UTB-JL—FET with
Lg = 3 nm in Si and Ge channels, respectively. UTB-JL—FET with Lg = 3 nm has
lower SS and DIBL than Lg = 1 nm. The SS is 84 mV/decade of Si pFET and
83 mV/decade of Si nFET, respectively. It is worth noting that we have not used
any strain engineering technique in our simulated device. The SS is 81 mV/decade
of Ge pFET and 79 mV/decade of Ge nFET, respectively because Ge has higher
channel mobility than Si channel. The Ge nFET has a 42% higher Isat than Si nFET,
and the Ge pFET has a 29% higher Isat than Si pFET.
Figure 8.23a, b plots the timing characteristics of a CMOS inverter and static
transfer characteristic curves of Si UTB-JL—FET with Lg = 1-nm 6T-SRAM cells,
respectively. Figure 8.23c, d plots the timing characteristics of an CMOS inverter and
Energy (ev)
0.0 0.0
-0.2 Ec -0.2 Ec
-0.4 -0.4
-0.6 -0.6
-0.8 -0.8
W 8 W 8
id 6 Lg id 6 Lg
t h t h
d Y 4 70 Y 4 70
d
(n ire 2 60 nm) (n ire 2
60
X( nm)
m ct X( 50
ion,
50 m ct
) io 0 40 ct ion, ) io 0 40
rect
n, dire n, l di
Vg = 0 V n nel Vg = 0.8 V nne
Vd = 0.8 V Cha Vd = 0.8 V Cha
Fig. 8.20 Conduction band energy (Ec) diagrams of UTB-JL—FET with Lg = 1 nm and
T = 1 nm in (a) off-state and (b) on-state. In off-state, a high Ec level at channel region and
depletion width of 2.5 nm at both source and drain side to block leakage current [4]
300 8 Extremely Scaled Si and Ge to Lg = 3 nm FinFETs …
Fig. 8.21 Id–Vg of UTB-JL—FET with channel thickness (T) is 1 nm, and gate length (Lg) is
1 nm for (a) silicon and (b) germanium channel [4]
pFET pFET
10 -4 Lg = 3 nm 10 -4 Lg = 3 nm
10 -5 10 -5
10 -6 Vs = -0.05 V Vd = 0.05 V 10 -6 Vs = -0.05 V Vd = 0.05 V
Fig. 8.22 Id–Vg of UTB-JL—FET with channel thickness (T) is 1 nm, and gate length (Lg) is
3 nm for (a) silicon and (b) germanium channel [4]
Voltage (V)
Voltage (V)
0.6
0.4 thl = tlh = 0.4
1.31 ps 1.61 ps Vdd = 0.8 V
0.2
0.2 SNM = 97 mV
0.0
-0.2 Silicon 0.0 Silicon
-0.4
0 25x10 -12 50x10 -12 75x10 -12 0.0 0.2 0.4 0.6 0.8
Time (sec) Voltage (V)
Voltage (V)
0.6
0.4 thl = tlh = 0.4
1.33 ps 1.49 ps Vdd = 0.8 V
0.2
0.2 SNM = 115 mV
0.0
-0.2 Germanium 0.0 Germanium
-0.4
0 25x10 -12 50x10 -12 75x10 -12 0.0 0.2 0.4 0.6 0.8
Time (sec) Voltage (V)
Fig. 8.23 a Timing characteristics of the input and output signals of a CMOS inverter for Si
UTB-JL—FET with Lg = 1 nm. b Static transfer characteristic curves of Si UTB-JL—
FET6T-SRAM cells. The definition of static noise margin (SNM) is the length of the side of
the largest square that can be embedded inside the butterfly curve. c Timing characteristics of the
input and output signals of an CMOS inverter for Ge UTB-JL—FET with Lg = 1 nm. d Static
transfer characteristic curves of Ge UTB-JL—FET 6T-SRAM cells [4]
reasonable SNM that can meet the industry requirements. Using focus ion beam
(FIB) or reactive-ion etching (RIE), this UTB recess channel structure can be
achieved in sub-5-nm CMOS technology nodes. And this device can integrate
high-k/metal gate by ALD and CMP. Finally, circuit performances reveal that
UTB-JL—FET can be used in advanced logic ICs applications.
Summary of this chapter: Advanced examples of Si and Ge FinFET with
Lg = 3 nm based on three different doping styles, such as inversion-mode (IM),
accumulation-mode (AC), and junctionless-mode (JL) FinFETs, are proposed in the
first half of this chapter, and the performances of all devices have been compared
and analyzed. The simulation results reveal that the electric properties of
inversion-mode, accumulation-mode, and junctionless-mode FinFETs are rather
similar to each other in ultra-fine nanoscale channels. The UTB-JL—FET with Lg
down to 1 nm has been proposed in the second half of this chapter. The simulation
results reveal that extreme scaling UTB-JL—FET still maintains excellent electric
properties. In addition, the characteristics of inverter and SRAM based on
302 8 Extremely Scaled Si and Ge to Lg = 3 nm FinFETs …
Voltage (V)
0.6
0.6
Voltage (V)
Fig. 8.24 a Timing characteristics of the input and output signals of a CMOS inverter for Si UTB-JL
—FET with Lg = 3 nm. b Static transfer characteristic curves of Si UTB-JL—FET6T-SRAM cells.
The definition of static noise margin (SNM) is the length of the side of the largest square that can be
embedded inside the butterfly curve. c Timing characteristics of the input and output signals of a
CMOS inverter for Ge UTB-JL—FET with Lg = 3 nm. d Static transfer characteristic curves of Ge
UTB-JL—FET 6T-SRAM cells [4]
References
1. S.D. Suk, M. Li, Y.Y. Yeoh, K.H. Yeo, J. K. Ha, H. Lim, H.W. Park, D.W. Kim, T.Y. Chung,
K.S. Oh, W.S. Lee, Characteristics of sub 5 nm tri-gate nanowire MOSFETs with single and
poly Si channels in SOI structure. VLSI Tech. Symp. 142 (2009)
2. S. Migita, Y. Morita, M. Masahara, H. Ota, Electrical performances of junctionless-FETs at the
scaling limit (Lch = 3 nm). Tech. Digest of IEDM, 8.6.1 (2012)
References 303
3. V. Thirunavukkarasu, Y.R. Jhan, Y.B. Liu, Y.C. Wu, Performance of inversion, accumulation,
and junctionless mode n-type and p-type bulk silicon FinFETs with 3-nm gate length. IEEE
Electr. Dev. Lett. 36, 645 (2015)
4. Y.R. Jhan, V. Thirunavukkarasu, C.P. Wang, Y.C. Wu, Performance evaluation of silicon and
germanium ultrathin body (1 nm) junctionless field-effect transistor with ultrashort gate length
(1 nm and 3 nm). IEEE Electr. Dev. Lett. 36, 654 (2015)
Appendix
Synopsys Sentaurus TCAD 2014 Version
Software Installation and Environmental
Settings
This is about the introduction of the latest Synopsys Sentaurus TCAD 2014 version
(http://www.synopsys.com/tools/tcad/Pages/default.aspx) Copyright © 2015
Synopsys, Inc. The later version installation is similar. This simulation software
(Sentaurus TCAD) can only be executed in Linux operating system, so this chapter
will start with instructions on how to establish a Linux operating system environ-
ment under Windows environment following the sequence as shown below:
1. Downloading and installation of VMware Workstation.
2. Installation steps of VMware Workstation.
3. Installation of Synopsys Sentaurus TCAD software.
Although the installation illustration has some Chinese characters, we explain in
English in all figure captions.
The first thing is about the recommendations for professional accessories of
simulation PC host:
platform. Users will be allow to user multiple operating systems, including Linux,
Windows or any other operating system on the same computer to execute appli-
cation programs without the need for rebooting.
2. Installation steps of VMware Workstation
Step 1: Select [Create a New Virtual Machine]. (copyright © 2015 VMware, Inc).
Step 3: Select the image of [CentOS-6.4]. This is because TCAD 2014 version can
only be executed in Linux operating system.
Step 5: Keep the original default value and directly go to the next step.
Step 6: Select the size of hard drive to be split for the Virtual Machine. The
recommended size is more than 100 GB.
Appendix: Synopsys Sentaurus TCAD 2014 Version Software … 309
Step 9: Select [processor]. This step is mainly for determine (1) How many cores of
CPU are to be assigned to this Virtual Machine; (2) How many threads per core are
to be assigned to this Virtual Machine? The partition in this step should be in
accordance with computer specification. A greater portion assigned to this Virtual
Machine will accelerate the simulation.
Step 11: Start [Virtual Machine] to begin with the installation of CentOS system.
Step 12: Press [Enter] upon entering this page. The user can install newest version
in CentOS website.
312 Appendix: Synopsys Sentaurus TCAD 2014 Version Software …
Step 14: Press [Next] in CentOS system to enter the installation setting menu.
Appendix: Synopsys Sentaurus TCAD 2014 Version Software … 313
Step 15: Setting the language of installation, in this installation using Chinese.
Step 21: Enter root password and confirm, then go next setp.
Step 22: Select [Use the available space] in the next step just to be safe, and then go
next setp.
Appendix: Synopsys Sentaurus TCAD 2014 Version Software … 317
Step 26: [Select Time]. This time must be consistent with the actual time of the
moment otherwise there might be malfunction. Or the synchronization via the
Internet can also be selected.
Appendix: Synopsys Sentaurus TCAD 2014 Version Software … 319
Step 27: Click select [Start kdump(E)] and then push complete bottom.
Step 28: The [Display] on the System Preference Menu can be selected to change
the screen resolution. The time on the upper right corner of the screen must be
changed to the local time, or TCAD will not function normally.
320 Appendix: Synopsys Sentaurus TCAD 2014 Version Software …
(1) installer_v3.1.tar.Z
(2) sentaurus_vj_2014.09_common.tar
(3) sentaurus_vj_2014.09_amd64.tar
All aforementioned files can be accessed via FTP connection to (1) Synopsys;
(2) National Center of High-Performance Computing of all countries. For example:
National Center for High-performance Computing (NCHC) of Taiwan at https://
www.nchc.org.tw/tw/, where three files will be saved in the same folder.
Appendix: Synopsys Sentaurus TCAD 2014 Version Software … 321
Step 30: Right click the desktop and select [open in Terminal] to access the terminal
and type [cd /home/lab203/tcad1] (© Synopsys, InC).
Step 39: Select Red Hat or SUSE in accordance with the version of Linux before
pressing [Next].
326 Appendix: Synopsys Sentaurus TCAD 2014 Version Software …
Step 40: This step should be changed to home to avoid any problem.
Step 41: The next step is for setting environmental variables. The first thing is to
switch to root account.
Appendix: Synopsys Sentaurus TCAD 2014 Version Software … 327
Step 42: Enter root as the ID and password and press Login, and then press Close
on the popped up window.
Step 44: Search for [profile] and then open this file.
Step 45: The two lines at the bottom should be added. The path for TCAD
installation should be entered as the path, and those following/bin will not be
changed. Press Save on the top bar after this step is completed.
Appendix: Synopsys Sentaurus TCAD 2014 Version Software … 329
Step 46: Leave the root account and return to the original account. Access the
terminal and type source/etc/profile.
Step 47: And then type export |grep LM_LICENSE_FILE and export |grep PATH
to verify the correctness of environmental variables.
330 Appendix: Synopsys Sentaurus TCAD 2014 Version Software …
Step 49: Select proper location for saved file, and it can be a self-defined folder.
And then press OK to start the simulation program.